Publications | Jorge A. Gómez-García

Thesis

2018

upm

Contributions to the Design of Automatic Voice Quality Analysis Systems using Speech Technologies

Jorge Andrés Gómez-García

Jan 2018

@thesis{gomez2018contributions,
  author = {Gómez-García, Jorge Andrés},
  doi = {10.20868/UPM.thesis.49565},
  institution = {Universidad Politécnica de Madrid},
  month = jan,
  title = {Contributions to the Design of Automatic Voice Quality Analysis Systems using Speech Technologies},
  year = {2018},
  url = {http://oa.upm.es/49565/},
}

2008

unal

Metodología para la caracterización de bioseñales empleando análisis de complejidad sobre espacio de embebimiento no uniforme

Jorge Andrés Gómez-García

Jan 2008

Bib HTML

@thesis{gomezmetodologia,
  author = {Gómez-García, Jorge Andrés},
  institution = {Universidad Nacional de Colombia},
  title = {Metodología para la caracterización de bioseñales empleando análisis de complejidad sobre espacio de embebimiento no uniforme},
  year = {2008},
  url = {https://repositorio.unal.edu.co/handle/unal/70145},
}

Journals

2024

Estimation of the Cyclopean Eye From Binocular Smooth Pursuit Tests

Elisa Luque-Buzo, Mehdi Bejani, Julián David Arias-Londoño, Jorge Andrés Gómez-García, Francisco Grandas-Pérez, and Juan Ignacio Godino-Llorente

IEEE Transactions on Cognitive and Developmental Systems, Dec 2024

Abs DOI

In binocular vision, the visual system combines images in the retina to generate a single perception, which triggers a sensorimotor process that forces the eyes to point to the same target. Thus, following a moving target, both eyes are expected to move synchronously following identical motor triggers but, in practise, significant differences between eyes are found due to the presence of certain artifacts and effects. Thus, a better indirect characterisation of the underlying neurological behaviour during eyes motion would require new automatic preprocessing methods applied to the eye tracking sequences for rendering the common and most significant movements of both eyes. To address this need, the present study proposes an automatic method for extracting the common components of the left and right eye motions from a set of Smooth Pursuit tests by applying an Independent Component Analysis. To do so, both sequences are decomposed into two independent latent components: the first, presumably correlated with the common motor triggering at the brain, while the second collects artifacts introduced during the recording process and small effects due to convergence deficits and eye dominance biases. The evaluations were carried out using data corresponding to 12 different smooth pursuit eye movements tests, which were collected using an infrared high-speed video-based eye tracking device from 41 parkinsonian patients and 47 controls. The results show that the automatic method can separate the aforementioned components in 99.50% of cases, extracting a latent component correlated with the common motor triggering at the brain, which we hypothesise is characterising the movements of the cyclopean eye. The estimated component could be used to simplify any other potential automatic analysis.
NeuroVoz: a Castillian Spanish corpus of parkinsonian speech

Janaína Mendes-Laureano, Jorge Andrés Gómez-García, Alejandro Guerrero-López, Elisa Luque-Buzo, Julián David Arias-Londoño, Francisco Grandas-Pérez, and Juan Ignacio Godino-Llorente

Scientific Data, Dec 2024

Abs DOI

The advancement of Parkinson’s Disease (PD) diagnosis through speech analysis is hindered by a notable lack of publicly available, diverse language datasets, limiting the reproducibility and further exploration of existing research. In response to this gap, we introduce a comprehensive corpus from 108 native Castilian Spanish speakers, comprising 55 healthy controls and 53 individuals diagnosed with PD, all of whom were under pharmacological treatment and recorded in their medication-optimized state. This unique dataset features a wide array of speech tasks, including sustained phonation of the five Spanish vowels, diadochokinetic tests, 16 listen-and-repeat utterances, and free monologues. The dataset emphasizes accuracy and reliability through specialist manual transcriptions of the listen-and-repeat tasks and utilizes Whisper for automated monologue transcriptions, making it the most complete public corpus of Parkinsonian speech, and the first in Castillian Spanish. NeuroVoz is composed by 2,903 audio recordings averaging 26.88 \pm 3.35 recordings per participant, offering a substantial resource for the scientific exploration of PD’s impact on speech. This dataset has already underpinned several studies, achieving a benchmark accuracy of 89% in PD speech pattern identification, indicating marked speech alterations attributable to PD. Despite these advances, the broader challenge of conducting a language-agnostic, cross-corpora analysis of Parkinsonian speech patterns remains an open area for future research. This contribution not only fills a critical void in PD speech analysis resources but also sets a new standard for the global research community in leveraging speech as a diagnostic tool for neurodegenerative diseases.
Detecting Postural Instability in Parkinson’s Disease from IMU-based Objective Measures

Kendal Smith, Diego Torricelli, Miguel González-Sánchez, Javier Ricardo Pérez-Sánchez, Elisa Luque-Buzo, Francisco Grandas, Juan Miguel Marín, and Jorge Andrés Gómez-García

Dec 2024

Abs DOI

Parkinson’s Disease (PD) is a progressive neurological disorder that affects both motor and nonmotor functions. A critical aspect of PD is postural instability, which is associated with a higher risk of falls and significant decline in mobility. The onset of postural instability is frequently identified through subjective clinical assessments such as the pull test, which aligns with the transition from Stage II to Stage III on the Hoehn and Yahr (H&Y) scale, a measure of general motor impairment. This study aims to pinpoint this transition by providing an objective quantification of the pull test, to prevent falls and manage severe motor symptoms via early detection. Sixty objective metrics were characterised from acceleration data using inertial measurement units (IMU) during pull tests conducted with 30 PD patients to differentiate between H&Y Stages II and III. Significant differences in feature distributions were identified, with two features achieving a ROC AUC of 0.95: chest to lumbar maximum acceleration ratio, and power ratio of 8-12 Hz versus total acceleration power. This paper demonstrates the ability of certain IMU-derived features to distinguish between H&Y Stages II and III, presenting more straightforward and objective measurements compared to conventional pull test metrics. Furthermore, the study explores the inconsistencies in the administration of the pull test by different examiners and shows that even with these inconsistencies, objective measures can still effectively differentiate between stages. To the best of our knowledge, this work represents the first extensive attempt to objectively quantify the pull test for this type of classification.

2023

Baseline Wander Removal Applied to Smooth Pursuit Eye Movements From Parkinsonian Patients

Mehdi Bejani, Elisa Luque-Buzo, Arsen Burlaka-Petrash, Jorge Andrés Gómez-García, Julián David Arias-Londoño, Francisco Grandas-Pérez, Jesus Grajal, and Juan Ignacio Godino-Llorente

IEEE Access, Dec 2023

Abs DOI

Prior studies aiming to parametrize the sequences obtained from the Smooth Pursuit Eye Movements (SPEM) of patients with Parkinson’s disease are based on the manual extraction of cues of interest. This is because methods to automatically extract the relevant information are complex to implement and are constrained, in part, by the appearance of a baseline wander (BW). Thus, new methods are required for preprocessing the SPEM sequences to make the potential parameterisation procedures much more robust, removing the aforementioned BW. In this respect, the present study compares different BW removal methods applied to SPEM sequences based on several objective evaluation metrics. At the same time, it proposes a set of guidelines to estimate the ground truth that is required for comparison purposes. Data were collected using a high-speed video-based eye-tracking device. 52 patients and 60 controls and 12 young participants were enrolled in the study. The ground truth required to compare the different BW removal techniques was manually delineated according to a predefined protocol. Seven methods were developed to remove the BW, and four objective metrics were used to evaluate the results. According to the results, a method based on the Empirical Wavelet Transform provided the best performance removing the BW. Furthermore, the objective and subjective results show that potential asymmetries between left and right eye movements are solved by removing the BW. Regardless of the techniques used, BW removal is revealed to be a crucial step for any autonomous SPEM processing tool.

2021

BSPC

On the design of automatic voice condition analysis systems. Part III: review of acoustic modelling strategies

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, Julián David Arias-Londoño, and Juan Ignacio Godino-Llorente

Biomedical Signal Processing and Control, Apr 2021

Abs DOI

This is the third of a three-part series devoted to review the current state of the art of automatic voice condition analysis systems. A direct continuation to “On the design of automatic voice condition analysis systems. Part I: review of concepts and an insight to the state of the art” and to “On the design of automatic voice condition analysis systems. Part II: review of speaker recognition techniques and study on the effects of different variability factors” already published in this journal. The goal of this paper is to compile the most significant parameterisation approaches used in the literature for automatic voice condition analysis systems, along with a critical discussion about their usefulness, providing the user with a comprehensive review of the most important techniques used for acoustic modelling in the field. The paper presents the mathematical formulation and physical interpretation of a series of perturbation and fluctuation parameters, noise features, complexity based parameters, modulation spectra, morphological parameters, and spectral-cepstral coefficients; and is complemented with a library written in MATLAB®, which has been made available to the readers in an online software repository.
BSPC

Advances in Parkinson’s Disease detection and assessment using voice and speech: A review of the articulatory and phonatory aspects

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Julián David Arias-Londoño, Najim Dehak, and Juan Ignacio Godino-Llorente

Biomedical Signal Processing and Control, Apr 2021

Abs DOI

Parkinson’s Disease (PD) affects speech in the form of dysphonia and hypokinetic dysarthria. Multiple studies have evaluated PD’s influence on different aspects of speech, showing differences between speakers with and without PD. Most recent studies are focused on the proposal of new automatic and objective tools to help in the diagnosis and severity assessment. This comprehensive review identifies the most common features and machine learning techniques employed in automatically detecting and assessing the severity of PD using phonatory and articulatory aspects of speech and voice. We discuss their discriminant properties and literature findings as well as identify common methodological issues that can potentially bias results. The objective is to provide a broad overview of these methods, their advantages and disadvantages, and to identify the most promising methodologies to be explored in future works. We conclude that there is clear evidence that the articulatory and phonatory aspects of speech and voice are relevant for the automatic detection and severity assessment of PD. However, there is no standard methodology sufficiently validated in a clinical trial, and further research is required, especially to develop larger corpora and identify new objective biomarkers.

2020

Analysis of the Effects of Supraglottal Tract Surgical Procedures in Automatic Speaker Recognition Performance

Laureano Moro-Velázquez, Estefania Hernandez-Garcia, Jorge Andrés Gómez-García, Juan Ignacio Godino-Llorente, and Najim Dehak

IEEE/ACM Transactions on Audio, Speech, and Language Processing, Apr 2020

Abs DOI

This article evaluates the impact in the performance of state-of-The-Art automatic speaker recognition schemes of three surgical procedures modifying the supraglottal tract structures of speakers. To do so, a new corpus (Cuco) was recorded, containing the speech of 107 speakers before and after surgery. Speakers were divided into four groups depending on the type of surgery: Tonsillectomy, functional endoscopy sinus surgery (FESS), septoplasty, and controls. The analyzed speaker recognition schemes were i-vectors, i-vectors with supervised Universal Background Model, i-vectors employing Time-delay Deep Neural Networks and x-vectors. In all cases, probabilistic linear discriminant analysis was employed in the back-end. Results show changes in the speech of patients who underwent tonsillectomy or FESS after surgery in contrast to controls or patients who had a septoplasty, where not significant variations are observed. These changes increase the Equal Error Rate (EER) of the analyzed speaker recognition schemes for the septoplasty and FESS groups when employing enrollment data recorded before the surgery. Moreover, surgery has a similar influence in the speech of female and male speakers with respect to the analyzed schemes. In consequence, results suggest that it is advisable to update the speaker’s enrollment speech after three months following supraglottal tract surgery to ensure that the effects of the operation and post-operative recovery period do not influence the performance of the automatic speaker recognition systems.
Artificial Intelligence Applied to Chest X-Ray Images for the Automatic Detection of COVID-19. A Thoughtful Evaluation Approach

Julián David Arias-Londoño, Jorge Andrés Gómez-García, Laureano Moro-Velázquez, and Juan Ignacio Godino-Llorente

IEEE Access, Apr 2020

Abs DOI

Current standard protocols used in the clinic for diagnosing COVID-19 include molecular or antigen tests, generally complemented by a plain chest X-Ray. The combined analysis aims to reduce the significant number of false negatives of these tests and provide complementary evidence about the presence and severity of the disease. However, the procedure is not free of errors, and the interpretation of the chest X-Ray is only restricted to radiologists due to its complexity. With the long term goal to provide new evidence for the diagnosis, this paper presents an evaluation of different methods based on a deep neural network. These are the first steps to develop an automatic COVID-19 diagnosis tool using chest X-Ray images to differentiate between controls, pneumonia, or COVID-19 groups. The paper describes the process followed to train a Convolutional Neural Network with a dataset of more than 79, 500 X-Ray images compiled from different sources, including more than 8, 500 COVID-19 examples. Three different experiments following three preprocessing schemes are carried out to evaluate and compare the developed models. The aim is to evaluate how preprocessing the data affects the results and improves its explainability. Likewise, a critical analysis of different variability issues that might compromise the system and its effects is performed. With the employed methodology, a 91.5% classification accuracy is obtained, with an 87.4% average recall for the worst but most explainable experiment, which requires a previous automatic segmentation of the lung region.
Multimodal and Multi-Output Deep Learning Architectures for the Automatic Assessment of Voice Quality Using the GRB Scale

Julián David Arias-Londoño, Jorge Andrés Gómez-García, and Juan Ignacio Godino-Llorente

IEEE Journal of Selected Topics in Signal Processing, Feb 2020

Abs DOI

This article addresses the automatic assessment of voice quality according to the GRB scale, based on the use of a variety of deep learning architectures for prediction purposes. The proposed architectures are multimodal, because they employ multiples sources of information; and also multi-output, because they simultaneously predict all the traits of the GRB scale. A feature engineering approach is followed, based on the use of deep neural networks and a set of well-established features such as MFCC, perturbation and complexity characteristics. Likewise, a representation learning is considered, using convolutional neural networks feed on modulation spectra extracted from voices. Finally, diverse loss functions are also investigated, including two surrogate ordinal classification, a conventional weighed categorical cross-entropy, and a mean square error function. Experiments are carried out in a dataset containing registers of the sustained phonation of three vowels. The best deep learning architecture provides a relative performance improvement of 6.25% for G, 14.1% for R and 18.1% for B, in comparison with recently published results using the same dataset.

2019

BSPC

On the design of automatic voice condition analysis systems. Part I: Review of concepts and an insight to the state of the art

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, and Juan Ignacio Godino-Llorente

Biomedical Signal Processing and Control, May 2019

Abs DOI

This is the first of a two-part series devoted to review the current state of the art of automatic voice condition analysis systems. The goal of this paper is to provide to the scientific community and to newly comers to the field of automatic voice condition analysis a resource that presents introductory concepts, a categorisation of different aspects of voice pathology and a systematic literature review that describes the methodologies and methods that are mostly used in these systems. To this end, pathological voice is firstly described in terms of perceptual characteristics and its relationship with physiological phenomena. Then, a prototypical automatic voice condition analysis system is described, discussing each one of its constituting parts and presenting an in-depth literature review about the methodologies that are typically employed. Finally, a discussion about some variability factors that affect the performance of these systems is presented.
Emulating the perceptual capabilities of a human evaluator to map the GRB scale for the assessment of voice disorders

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, Janaína Mendes-Laureano, Germán Castellanos-Domínguez, and Juan Ignacio Godino-Llorente

Engineering Applications of Artificial Intelligence, Jun 2019

Abs DOI

This paper presents the design of an automatic voice quality analysis system for the assessment of voice pathologies, which emulates the perceptual capabilities of a human evaluator according the GRB scale. For this purpose, a novel methodology based on multiple sets of characteristics, ordinal classification and Gaussian regression is proposed. In particular, a reduced subset of characteristics is identified, and the regressor is used to convert the discrete perceptual scale to a continuum, more in agreement to the nature of the problem under study. The robustness of the system is evaluated in several cross-dataset experiments. Similarly, a clinical evaluation of the predictions provided by the system is carried out. Results indicate that the proposed methodology is proficient in modelling the perceptual capabilities of the human evaluator. They also show that it is possible to extend the GRB scale to a continuum through regression techniques while maintaining the consistency of the results. On average, the deviation between the labels assessed by the expert and the ones provided by the system is of about 0.5 units (in a scale from 0 to 3) for G and B, and of 0.7 units for R. Similarly, the deviation of the labels predicted by the system in the clinical assessment trials is about 0.3 units for G, 0.4 units for B, and 0.5 units for R.
Phonetic relevance and phonemic grouping of speech in the automatic detection of Parkinson’s Disease

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Juan Ignacio Godino-Llorente, Francisco Grandas-Pérez, Stephanie Shattuck-Hufnagel, Virginia Yagüe-Jimenez, and Najim Dehak

Scientific Reports, Dec 2019

Abs DOI

Literature documents the impact of Parkinson’s Disease (PD) on speech but no study has analyzed in detail the importance of the distinct phonemic groups for the automatic identification of the disease. This study presents new approaches that are evaluated in three different corpora containing speakers suffering from PD with two main objectives: to investigate the influence of the different phonemic groups in the detection of PD and to propose more accurate detection schemes employing speech. The proposed methodology uses GMM-UBM classifiers combined with a technique introduced in this paper called phonemic grouping, that permits observation of the differences in accuracy depending on the manner of articulation. Cross-validation results reach accuracies between 85% and 94% with AUC ranging from 0.91 to 0.98, while cross-corpora trials yield accuracies between 75% and 82% with AUC between 0.84 and 0.95, depending on the corpus. This is the first work analyzing the generalization properties of the proposed approaches employing cross-corpora trials and reaching high accuracies. Among the different phonemic groups, results suggest that plosives, vowels and fricatives are the most relevant acoustic segments for the detection of PD with the proposed schemes. In addition, the use of text-dependent utterances leads to more consistent and accurate models.
BSPC

A forced gaussians based methodology for the differential evaluation of Parkinson’s Disease by means of speech processing

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Juan Ignacio Godino-Llorente, Jesús Villalba, Jan Rusz, Stephanie Shattuck-Hufnagel, and Najim Dehak

Biomedical Signal Processing and Control, Feb 2019

Abs DOI

Literature evidences the existence of hypokinetic dysarthria in parkinsonian patients and, consequently, the objective characterization of the dysarthric signs associated to the articulatory aspect of speech can be used to detect Parkinson’s Disease (PD) providing clinicians with new tools to support the clinical diagnosis. However, no work has analyzed in detail the importance of the different phonemes in the automatic detection of PD from the speech. This work proposes new approaches for this detection by using new classification schemes that allow to compare independently the different phonetic units of patients and controls employed during several speech tasks. Three different parkinsonian corpora were used allowing cross-validation and cross-corpora trials. The results of cross-validation trials (k-folds) provided accuracies between 81% and 94%, with AUC between 0.87 and 0.97 depending on the corpus, while cross-corpora trials yielded accuracies between 66% and 76% with AUC between 0.76 and 0.87. These results suggest that PD affects to the articulatory sequence as a whole, influencing more clearly phonetic units requiring a higher narrowing of the vocal tract. Additionally, text-dependent utterances are considered as the recommended speech task for the detection of PD in this type of schemes as these allow to compare more precisely the phonetic units of patients and controls. Lastly, this work discusses the existence of a glass ceiling in the accuracy of the systems for the automatic detection of PD using speech, concluding that this is below 95% for most of the cases.
BSPC

On the design of automatic voice condition analysis systems. Part II: Review of speaker recognition techniques and study on the effects of different variability factors

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, and Juan Ignacio Godino-Llorente

Biomedical Signal Processing and Control, Feb 2019

Abs DOI

This is the second of a two-part series devoted to the automatic voice condition analysis of voice pathologies, being a direct continuation to the paper On the design of automatic voice condition analysis systems. Part I: review of concepts and an insight to the state of the art. The aim of this study is to examine several variability factors affecting the robustness of systems that automatically detect the presence of voice pathologies by means of audio registers. Multiple experiments are performed to test out the influence of the speech task, extralinguistic aspects (such as sex), the acoustic features and the classifiers in their performance. Some experiments are carried out using state-of-the-art classification methodologies often employed in speaker recognition. In order to evaluate the robustness of the methods, testing is repeated across several corpora with the aim to create a single system integrating the conclusions obtained previously. This system is later tested under cross-dataset scenarios in an attempt to obtain more realistic conclusions. Results identify a reduced subset of relevant features, which are used in a hierarchical-like scenario incorporating information of different speech tasks. In particular, for the experiments carried out using the Saarbrüecken voice dataset, the area under the ROC curve of the system reached 0.88 in an intra-dataset setting and ranged from 0.82 to 0.94 in cross-dataset scenarios. These results let us open a discussion about the suitability of these techniques to be transferred to the clinical setting.

2018

Analysis of speaker recognition methodologies and the influence of kinetic changes to automatically detect Parkinson’s Disease

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Juan Ignacio Godino-Llorente, Jesús Villalba, Rafael Orozco-Arroyave, and Najim Dehak

Applied Soft Computing, Jan 2018

Abs DOI

The diagnosis of Parkinson’s Disease is a challenging task which might be supported by new tools to objectively evaluate the presence of deviations in patient’s motor capabilities. To this respect, the dysarthric nature of patient’s speech has been exploited in several works to detect the presence of this disease, but none of them has deeply studied the use of state-of-the-art speaker recognition techniques for this task. In this paper, two classification schemes (GMM-UBM and i-Vectors-GPLDA) are employed separately with several parameterization techniques, namely PLP, MFCC and LPC. Additionally, the influence of the kinetic changes, described by their derivatives, is analysed. With the proposed methodology, an accuracy of 87% with an AUC of 0.93 is obtained in the optimal configuration. These results are comparable to those obtained in other works employing speech for Parkinson’s Disease detection and confirm that the selected speaker recognition techniques are a solid baseline to compare with future works. Results suggest that Rasta-PLP is the most reliable parameterization for the proposed task among all the tested features while the two employed classification schemes perform similarly. Additionally, results confirm that kinetic changes provide a substantial performance improvement in Parkinson’s Disease automatic detection systems and should be considered in the future.

2017

Towards the identification of Idiopathic Parkinson’s Disease from the speech. New articulatory kinetic biomarkers

Juan Ignacio Godino-Llorente, Stephanie Shattuck-Hufnagel, Jeung-Yoon Choi, Laureano Moro-Velázquez, and Jorge Andrés Gómez-García

PLOS ONE, Dec 2017

Abs DOI

Although a large amount of acoustic indicators have already been proposed in the literature to evaluate the hypokinetic dysarthria of people with Parkinson’s Disease, the goal of this work is to identify and interpret new reliable and complementary articulatory biomarkers that could be applied to predict/evaluate Parkinson’s Disease from a diadochokinetic test, contributing to the possibility of a further multidimensional analysis of the speech of parkinsonian patients. The new biomarkers proposed are based on the kinetic behaviour of the envelope trace, which is directly linked with the articulatory dysfunctions introduced by the disease since the early stages. The interest of these new articulatory indicators stands on their easiness of identification and interpretation, and their potential to be translated into computer based automatic methods to screen the disease from the speech. Throughout this paper, the accuracy provided by these acoustic kinetic biomarkers is compared with the one obtained with a baseline system based on speaker identification techniques. Results show accuracies around 85% that are in line with those obtained with the complex state of the art speaker recognition techniques, but with an easier physical interpretation, which open the possibility to be transferred to a clinical setting.
Análisis de la influencia de las técnicas de compresión de voz en la detección de anomalías vocales

Lina María Sepúlveda Cano, Jhon Jair Quiza Montealegre, and Jorge Andrés Gómez-García

Revista Ingenierías Universidad de Medellín, Dec 2017

DOI

2016

Voice Pathology Detection Using Modulation Spectrum-Optimized Metrics

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, and Juan Ignacio Godino-Llorente

Frontiers in Bioengineering and Biotechnology, Jan 2016

Abs DOI

There exist many acoustic parameters employed for pathological assessment tasks, which have served as tools for clinicians to distinguish between normophonic and patho-logical voices. However, many of these parameters require an appropriate tuning in order to maximize its efficiency. In this work, a group of new and already proposed modulation spectrum (MS) metrics are optimized considering different time and frequency ranges pursuing the maximization of efficiency for the detection of pathological voices. The optimization of the metrics is performed simultaneously in two different voice databases in order to identify what tuning ranges produce a better generalization. The experiments were cross-validated so as to ensure the validity of the results. A third database is used to test the optimized metrics. In spite of some differences, results indicate that the behavior of the metrics in the optimization process follows similar tendencies for the tuning databases, confirming the generalization capabilities of the proposed MS metrics. In addition, the tuning process reveals which bands of the modulation spectra have relevant information for each metric, which has a physical interpretation respecting the phonatory system. Efficiency values up to 90.6% are obtained in one tuning database, while in the other, the maximum efficiency reaches 71.1%. Obtained results also evidence a separability between normophonic and pathological states using the proposed metrics, which can be exploited for voice pathology detection or assessment.
An insight to the automatic categorization of speakers according to sex and its application to the detection of voice pathologies: A comparative study

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, Juan Ignacio Godino-Llorente, and Germán Castellanos-Domínguez

Revista Facultad de Ingeniería Universidad de Antioquia, Jan 2016

DOI
Spectrum Sensing Framework based on Blind Source Separation for Cognitive Radio Environments

Lina María Sepúlveda Cano, Jhon Jair Quiza Montealegre, Camilo Gil Taborda, and Jorge Andrés Gómez-García

Revista Ingenierías Universidad de Medellín, Jan 2016

DOI

2015

Modulation Spectra Morphological Parameters: A New Method to Assess Voice Pathologies according to the GRBAS Scale

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Juan Ignacio Godino-Llorente, and Gustavo Andrade-Miranda

BioMed Research International, Jan 2015

Abs DOI

Disordered voices are frequently assessed by speech pathologists using perceptual evaluations. This might lead to problems caused by the subjective nature of the process and due to the influence of external factors which compromise the quality of the assessment. In order to increase the reliability of the evaluations, the design of automatic evaluation systems is desirable. With that in mind, this paper presents an automatic system which assesses the Grade and Roughness level of the speech according to the GRBAS perceptual scale. Two parameterization methods are used: one based on the classic Mel-Frequency Cepstral Coefficients, which has already been used successfully in previous works, and other derived from modulation spectra. For the latter, a new group of parameters has been proposed, named Modulation Spectra Morphological Parameters : MSC, DRB, LMR, MSH, MSW, CIL, PALA, and RALA. In methodology, PCA and LDA are employed to reduce the dimensionality of feature space, and GMM classifiers to evaluate the ability of the proposed features on distinguishing the different levels. Efficiencies of 81.6% and 84.7% are obtained for Grade and Roughness, respectively, using modulation spectra parameters, while MFCCs performed 80.5% and 77.7%. The obtained results suggest the usefulness of the proposed Modulation Spectra Morphological Parameters for automatic evaluation of Grade and Roughness in the speech.
An automatic method to detect and track the glottal gap from high speed videoendoscopic images

Gustavo Andrade-Miranda, Juan Ignacio Godino-Llorente, Laureano Moro-Velázquez, and Jorge Andrés Gómez-García

BioMedical Engineering OnLine, Jan 2015

DOI

2014

Non uniform Embedding based on Relevance Analysis with reduced computational complexity: Application to the detection of pathologies from biosignal recordings

Jorge Andrés Gómez-García, Juan Ignacio Godino-Llorente, and Germán Castellanos-Domínguez

Neurocomputing, May 2014

Abs DOI

Nonlinear analysis tools for studying and characterizing the dynamics of physiological signals have gained popularity, mainly because tracking sudden alterations of the inherent complexity of biological processes might be an indicator of altered physiological states.Typically, in order to perform an analysis with such tools, the physiological variables that describe the biological process under study are used to reconstruct the underlying dynamics of the biological processes. For that goal, a procedure called time-delay or uniform embedding is usually employed. Nonetheless, there is evidence of its inability for dealing with non-stationary signals, as those recorded from many physiological processes.To handle with such a drawback, this paper evaluates the utility of non-conventional time series reconstruction procedures based on non uniform embedding, applying them to automatic pattern recognition tasks. The paper compares a state of the art non uniform approach with a novel scheme which fuses embedding and feature selection at once, searching for better reconstructions of the dynamics of the system. Moreover, results are also compared with two classic uniform embedding techniques. Thus, the goal is comparing uniform and non uniform reconstruction techniques, including the one proposed in this work, for pattern recognition in biomedical signal processing tasks. Once the state space is reconstructed, the scheme followed characterizes with three classic nonlinear dynamic features (Largest Lyapunov Exponent, Correlation Dimension and Recurrence Period Density Entropy), while classification is carried out by means of a simple k-nn classifier. In order to test its generalization capabilities, the approach was tested with three different physiological databases (Speech Pathologies, Epilepsy and Heart Murmurs).In terms of the accuracy obtained to automatically detect the presence of pathologies, and for the three types of biosignals analyzed, the non uniform techniques used in this work lightly outperformed the results obtained using the uniform methods, suggesting their usefulness to characterize non-stationary biomedical signals in pattern recognition applications. On the other hand, in view of the results obtained and its low computational load, the proposed technique suggests its applicability for the applications under study. © 2013 Elsevier B.V.

2010

Análisis de la tolerancia al ruido de características basadas en dinámica no lineal sobre señales fonocardiográficas

Carolina Ospina-Aguirre, Jorge Andrés Gómez-García, Edilson Delgado-Trejos, and Germán Castellanos-Domínguez

TecnoLógicas, Dec 2010

Abs DOI

Una propiedad muy deseable en sistemas de detección automática de patologías es la robustez ante el ruido, tal que su presencia no afecte notablemente la habilidad de detectar patologías. El presente trabajo estudia la capacidad discriminatoria que características de la dinámica no lineal, en particular, el Exponente de Hurst, el Máximo Exponente de Lyapunov y la Dimensión de Correlación, puedan proveer en la detección de soplos cardíacos usando señales fonocardiográficas contaminadas con diferentes niveles de ruido. A su vez se hacen comparaciones con características obtenidas a partir de representaciones tiempo frecuencia. Los resultados muestran la fortaleza de las características basadas en dinámica no lineal para tareas de clasificación de estados funcionales, aún sobre señales con altos niveles de ruido.

2008

Conferences

2024

Estabilidad de la marcha: estado del arte de las métricas actuales

Adriana Torres-Pardo, Carlotta Mummolo, David Rodríguez-Cianca, Jorge Andrés Gómez-García, Juan C. Moreno, and Diego Torricelli

In Actas del Simposio de Robótica, Bioingeniería y Visión por Computador, 2024
Gait Analysis of Electromyographic Spectral Differences in Stroke Survivors and Healthy Controls

Marina Algaba-Vidoy, Jorge Andrés Gómez-García, Filipe Oliveira-Barroso, Francisco Molina-Rueda, Diego Torricelli, and Juan C. Moreno

In Advances in Digital Health and Medical Bioengineering. EHB 2023. IFMBE Proceedings,, 2024

DOI
Differences in gait electromyographic activity in stroke and healthy subjects using Discrete Wavelet Transform

Marina Algaba-Vidoy, Jorge Andrés Gómez-García, Filipe Oliveira-Barroso, Enrique Navarro-Cabello, Francisco Molina-Rueda, Diego Torricelli, and Juan C. Moreno

In International Society of Electrophysiology and Kinesiology Congress (ISEK), 2024

Abs

Hemiplegic gait is a common outcome of stroke. Hence, to assess asymmetric control of poststroke gait, this work analyzed time-frequency domain (TFD) of surface electromyography (sEMG). Methodology: 9 chronic stroke patients (SP) and 10 healthy controls (HC) were compared using Discrete Wavelet Transform (DWT) to study the electrophysiology underlying tibialis anterior (TA) activity. DWT was applied to filtered sEMG signals during 10 walking trials. Percent of energy within frequency bands (25Hz bandwidth, 25 to 300Hz) was compared in stance (ST) and swing (SW) phases using Wilcoxon’s test (dominant vs. nondominant in HC, paretic vs. nonparetic in SP) and Mann-Whitney U-test (paretic, nonparetic vs. HC). Results: The energy at each frequency band may reflect the number and type of recruited motor units (MUs). For dominant vs. nondominant HC, a marked distribution along gait was found: low-frequency MUs prevailed in ST (25-50Hz), while slow (50-75Hz) and fast (200-225Hz) did in SW for effective foot clearance. Yet, differences in HC were observed, probably due to subjects’ variability or other effects (i.e., nondominant required more energy for the same task in ST). This physiological distribution was altered in SP. For paretic vs. HC, paretic ST had lower energy. In SW, it increased at 25-50Hz, but decreased at 75-100 and 150-300Hz. This overall reduction explains the typical TA weakness in stroke. Energy changes in bands suggest alterations in MUs recruitment: 50-75Hz MUs that sustain HC ST were replaced by slower MUs in paretic (25-50Hz). Also, faster MUs were reduced in paretic ST (75-125, 175-275Hz) and SW (75-100, 150-300Hz). For nonparetic vs. HC, energy was lower in 25-50Hz, but it increased in 75-99, 125-150 and 175-200Hz, likely because of nonparetic mechanisms that compensate the inability to activate the affected TA. During SW, nonparetic showed a rise in low frequencies (25-50, 50-75Hz) and a reduction in 175-250Hz, contrary to the HC pattern of MUs during SW. For paretic vs. nonparetic, energy in paretic was lower in all bands. In SW, energy in paretic increased in 25-50 and 225-300Hz bands and was reduced in 50-75 and 150-200Hz. Thus, paretic relies on the slowest (25-50Hz) and fastest (225-300Hz) MUs, rather than the 50-75Hz (HC and nonparetic) or 200-225Hz (HC) and 150-174Hz (nonparetic) ones. Conclusion: DWT of TA showed altered spectral attributes in stroke gait. The energy distribution in HC was modified in paretic and nonparetic, which may indicate a pathological alteration in the ratio of recruited MUs of each class during gait. In paretic SW this change is pronounced, possibly due to the occurrence of foot drop. Energy distribution characterized poststroke gait and asymmetries of dominant vs. nondominant and paretic vs. nonparetic. In the future, these findings will be compared with Continuous Wavelet Transform and new features will be extracted from TFD to provide a deeper description of neuromuscular altered mechanisms.
Applicability of Gait Indices in Parkinson’s Disease: A Review on Clinical, Methodological and Experimental Challenges

Nicolás Eugenio Gómez-Suárez, Jorge Andrés Gómez-García, José Javier Serrano Olmedo, Juan C. Moreno, and Diego Torricelli

In Converging Clinical and Engineering Research on Neurorehabilitation V, 2024

Abs DOI

This paper reviews the use of various gait indices for assessing Parkinson’s disease gait. Selected indices include the Gait Deviation Index, the Gillette Gait Index, the Gait Profile Score, the Hip Flexor Index, the GDI-Kinetic, the Symmetry Index, and the...
Generative Models for Data Augmentation on Inertial Measurement Units Data Classification

Blanca Llauradó, José Fernando Núñez, Axel Romero, Jamie Arjona, Javier Béjar, Jorge Andrés Gómez-García, and Diego Torricelli

Sep 2024

DOI
Is IMU-Based Center of Mass Estimation Reliable? Preliminary Study in Perturbed Conditions

Adriana Torres-Pardo, Carlota Trigo, Jorge Andrés Gómez-García, Juan C. Moreno, Stephanie L. Carey, and Diego Torricelli

In Converging Clinical and Engineering Research on Neurorehabilitation V, Sep 2024

Abs DOI

Falls, often occurring during walking, are a major health concern. Clinical stability evaluations are typically subjective and lack standardized methods. There is a need for objective, easy-to-implement, and standardized methods to evaluate stability in clinical...
Arm Swing Kinematic in Subjects with Early Parkinson’s Disease Compared with Healthy Controls When Walking on Unstable Terrain

Víctor Navarro-López, Simón García Diego-Martínez, María Carratalá-Tejada, Diego Fernández-Vázquez, Paula Molero-Mateo, Adriana Torres-Pardo, Marina Algaba-Vidoy, Jorge Andrés Gómez-García, İrem Akgün, Nicolás Eugenio Gómez-Suárez, Ainara Carpio-Chicote, Carlota Trigo, Diego Torricelli, and Francisco Molina-Rueda

In Converging Clinical and Engineering Research on Neurorehabilitation V, Sep 2024

Abs DOI

Reduced arm swing in Parkinson’s disease (PD) is considered a prodromal sign of the disease, being related to gait disturbances and an increased risk of falls. The aim of this study was to compare the arm swing kinematics of people with early PD and healthy...
An Algorithm for Gait Segmentation Using Surface Electromyography and Subsequence Dynamic Time Warping

Marina Algaba-Vidoy, Annemarie F. Laudanski, Filipe Oliveira-Barroso, Jorge Andrés Gómez-García, Diego Torricelli, and Juan C. Moreno

In Converging Clinical and Engineering Research on Neurorehabilitation V, Sep 2024

Abs DOI

Gait segmentation in surface electromyography (sEMG) studies is usually achieved through kinematics. However, there are occasions when this multimodal analysis cannot be implemented. This study presents an algorithm for automated gait segmentation using sEMG signals...
Multi-modal Technology-Based Assessment of Parkinson’s Disease: Technological Platform of the AI4HA Project

Francesca Lunardini, Jorge Andrés Gómez-García, Elisa Luque-Buzo, Karan Chugani, Teresa Rodríguez-Millán, Miguel Ángel Ferre-Bañón, Juan Ignacio Godino-Llorente, Aureli Soria-Frisch, Jesús Tornero, and Diego Torricelli

In Converging Clinical and Engineering Research on Neurorehabilitation V, Sep 2024

Abs DOI

This paper describes the technological platform of the AI4HA project, one of the first global efforts to develop a multimodal platform based on Artificial Intelligence (AI) for the analysis of Parkinson’s Disease (PD). It involves the multimodal analysis of...
Comparison of Muscle Synergies Between Parkinson’s Disease Patients and Healthy Controls During Walking Over a Mat

İrem Akgün, Marina Algaba-Vidoy, Adriana Torres-Pardo, Víctor Navarro-López, Diego Fernández-Vázquez, María Carratalá-Tejada, Francisco Molina-Rueda, Jorge Andrés Gómez-García, Juan C. Moreno, İlkşan Demirbüken, Filipe Oliveira-Barroso, and Diego Torricelli

In Converging Clinical and Engineering Research on Neurorehabilitation V, Sep 2024

Abs DOI

Patients who suffer from Parkinson’s disease (PD) usually present gait dysfunction. The analysis of muscle synergies has been used to unravel neural strategies underlying human locomotion. The main goal of this study was to compare muscle synergies between...
Clinically Meaningful Gait Feature Selection for Classification of Parkinson’s Disease

Ainara Carpio-Chicote, Jorge Andrés Gómez-García, Juan C. Moreno, Francisco Grandas-Pérez, and Diego Torricelli

In Converging Clinical and Engineering Research on Neurorehabilitation V, Sep 2024

Abs DOI

The diagnosis of Parkinson’s disease (PD) is challenging due to the overlap of its clinical signs with those of other neurodegenerative diseases. Gait analysis is a promising tool that can capture early and evolving motor abnormalities in PD. This study focuses...

2023

Is Lyapunov exponent a reliable metric to detect dynamic stability in Parkinson’s disease?

Adriana Torres-Pardo, Jorge Andrés Gómez-García, Nicolás Eugenio Gómez-Suárez, Adriana Muñoz-González, Miguel González-Sánchez, Francisco Grandas-Pérez, Juan C. Moreno, and Diego Torricelli

In 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), Jul 2023

Abs DOI

Idiopathic Parkinson’s disease (PD) is the second most common neurodegenerative disorder worldwide. It affects the nervous system, causing motor and non-motor symptomatology. However, its clinical diagnosis remains dependent on the expertise of clinicians, as perceptual clinical scales are often used. Gait stability is one of the most relevant motor signs in PD. Nonetheless, it is usually not reported or quantified, possibly due to its unclear meaning and the high variability of metrics used in the literature. This work aims to identify a reliable and objective indicator that clinicians can use to assess patients in realistic contexts. We focused on the Largest Lyapunov Exponent (LLE), being the most common metric used in previous research works to quantify gait stability. The short and long-term LLEs were calculated in a group of 34 healthy and 42 participants diagnosed with PD. The long-term LLE extracted from the chest, right arm and right foot sensors showed statistical differences between subjects with PD and healthy control (HC) subjects, showing that the HC subjects are more stable than PD patients, whereas the short-term LLE showed the opposite results. Further investigation is required to clarify the reliability of this metric to detect and rate gait stability in people affected with PD.Clinical Relevance - This study is the first step towards the identification of an objective methodology to assess gait stability in clinical settings. Achieving this goal will contribute to improve the understanding and support the diagnosis of gait disorders that cause gait stability problems.

2021

New tools for the differential evaluation of Parkinson’s disease using voice and speech processing

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Najim Dehak, and Juan Ignacio Godino-Llorente

In IberSPEECH 2021, Mar 2021

Abs DOI

Parkinson’s Disease (PD) is a neurodegenerative condition that affects the motor capabilities of individuals. Early detection can potentially contribute to slow its progression in a near future. Therefore, new objective and reliable tools are needed to support its diagnosis. Literature suggests that the patients’ speech can provide relevant information about the presence of the disease. In this study, five sets of experiments were carried out, each containing new approaches to detect the presence of the disease in the speech of idiopathic PD patients and control speakers from three parkinsonian corpora, two of them in the Spanish language. Different speech frame selection techniques are proposed, such as phonemic and acoustic landmark distillation, providing certain specific speech segments of interest to this work’s purposes. Multiple cepstral and spectral features were employed, along with several classification techniques based on Gaussian models and speaker embeddings. The best accuracy results in detecting PD with the proposed method-ologies reached values ranging from 85% to 94% with Area Under the Curve between 0.91 and 0.99, depending on the corpus. Results suggest that PD affects the movements related to all of the studied articulatory segmental groups but has a more evident influence in the consonants with a greater narrowing of the vocal tract, mainly plosives and fricatives. The new proposed methodologies demonstrate their ability to support PD’s diagnosis during a patient’s clinical assessment and are a step forward in PD’s speech-based diagnosis systems.

2020

Predicting UPDRS Scores in Parkinson’s Disease Using Voice Signals: A Deep Learning/Transfer-Learning-Based Approach

Julián David Arias-Londoño, and Jorge Andrés Gómez-García

Mar 2020

Abs DOI

In the last years, literature exhibits successful results in the automatic detection of Parkinson’s disease using voice/speech, especially for patients in medium or late stages of the disorder. By contrast, the prediction of the UPDRS scores -used to assess the severity of the disorder or the efficacy of treatments- has been shown to perform mostly poor. These results could be explained by the need of more complex machine learning models compared to the detection case, and the lack of large databases for properly training artificial intelligence models. To analyse possible solutions to these problems, this work will explore the potentiality of Deep Neural Network (DNN) and Convolutional Neural Network (CNN) models, along transfer learning approaches, for the automatic prediction of the UPDRS scores. Experiments are carried out using feature engineering and feature learning methodologies. In particular for feature engineering, a series of well-know features that are used to characterise vocal conditions are employed to train a DNN. Likewise, the feature learning approach is based on transformation of the input speech using Modulation spectra transformations to train a CNN, considering a transfer learning approach. For transfer learning, the networks will be trained using voice signals from patients of databases of organic and functional voice pathologies; following a network architecture that has been proven successful recently for voice quality assessment using the GRB scale. The approach includes the combination of feature learning and feature engineering approaches using a multimodal strategy. The fine-tuning procedure of the last layers in the second network will be carried out using two databases of PD patients. The results present insights about the potential of deep learning along with transfer learning strategies for the prediction of UPDRS score in parkinsonian speechs.
Approaches to Evaluate Parkinsonian Speech Using Artificial Models

Juan Ignacio Godino-Llorente, Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Jeung-Yoon Choi, Najim Dehak, and Stephanie Shattuck-Hufnagel

Mar 2020

Abs DOI

Our preliminary data and experiments show the potentiality of the artificial intelligence techniques to identify Parkinson’s disease and to assess its extent using the information extracted from the speech. Our conclusions are based on several prospective studies with different corpora of parkinsonian and control subjects modelled following different approaches, mainly grounded on articulatory aspects of the speech and on the use of probabilistic generative models inspired in those used in the field of speaker recognition. Results obtained are consistent among the different approaches and corpora, providing detection accuracies over 85%. This chapter compiles and compares the approaches we followed and the results we obtained in the last few years, and discusses some key aspects for future research in the field. The goal is not only to demonstrate the potentiality of the artificial intelligence techniques used, but also to extract from them significant knowledge that might be of interest for our understanding of the effects of the disease on the speech. To this respect a phonemic analysis based on the automatic techniques developed is carried out, suggesting important cues about the acoustic units that are more affected by the disease.

2019

Analysis of phonatory features for the automatic detection of Parkinson’s disease in two different corpora

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Najim Dehak, and Juan Ignacio Godino-Llorente

In Models and Analysis of Vocal Emissions for Biomedical Applications - 11th International Workshop, MAVEBA 2019, Mar 2019

Abs

In this study, several automatic detectors of Parkinson’s Disease (PD) based on phonatory aspects were analyzed employing two different corpora containing speech from speakers with PD. The features employed to characterize phonation were jitter, shimmer, noise measurements, complexity, modulation spectrum features and perceptual linear predictive coefficients. To differentiate between speakers with and without PD a gaussian mixture model classification scheme was used. Then, the approach providing the best results was combined with a scheme using articulatory information of the speech in order to assess the complementarity between phonatory and articulatory aspects in the automatic detection of PD. Cross-validation trials (k-folds) employing exclusively phonatory information provided accuracies between 64% and 71%, with AUC between 0.68 and 0.80 depending on the corpus. Results suggest that a combination of all the analyzed features with a PCA dimensionality reduction produce the best accuracy, AUC and sensibility. Also, results indicate that phonatory approaches tend to be less accurate in PD detection than other articulatory approaches proposed in previous studies. Finally, results suggest the complementarity between the studied articulatory and phonatory approaches is low.

2018

ByoVoz Automatic Voice Condition Analysis System for the 2018 FEMH Challenge

Julián David Arias-Londoño, Jorge Andrés Gómez-García, Laureano Moro-Velázquez, and Juan Ignacio Godino-Llorente

In 2018 IEEE International Conference on Big Data (Big Data), Dec 2018

Abs DOI

This paper presents the methods and results used by the ByoVoz team for the design of an automatic voice condition analysis system, which was submitted to the 2018 Far East Memorial Hospital voice data challenge. The proposed methodology is based on a cascading scheme that firstly discriminates between pathological and normophonic voices, and then identifies the type of disorder. By using diverse feature selection techniques, a subset of complexity, spectral/cepstral and perturbation characteristics were identified for the proposed tasks. Then, several generative classification methodologies based on Gaussian Mixture Models and Gradient Boosting were employed to provide decisions about the input voices in the binary classification, and using one-vs-one classification systems based on Random Forests for the categorization according to the type of disorder. By using a 4-folds cross-validation approach on the training partition a sensitivity=0.93 and specificity=0.74 was obtained. Similarly, an unweighted average recall of 0.63 and an accuracy of 66% was obtained for the identification task. Using the scoring metric proposed in the challenge the final resulting score is of 0.77. When the testing partition was considered, a sensitivity of 0.92, specificity of 0.54, unweighted average recall of 0.61 and a score of 0.72 were obtained. This was the fifth best result in the competition.
Study of the Automatic Detection of Parkison’s Disease Based on Speaker Recognition Technologies and Allophonic Distillation

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, Juan Ignacio Godino-Llorente, Jan Rusz, Sabine Skodda, Rafael Orozco-Arroyave, Elmer Noth, and Najim Dehak

In 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Jul 2018

Abs DOI

The use of new tools to detect Parkinson’s Disease (PD) from speech articulatory movements can have a considerable impact in the diagnosis of patients. In this study, a novel approach involving speaker recognition techniques with allophonic distillation is proposed and tested separately in four parkinsonian speech databases (205 patients and 186 controls in total). This new scheme provides values between 72% and 94% of accuracy in the automatic detection of PD, depending on the database, and improvements up to 9% respect to baseline techniques. Results not only point towards the importance of the segmentation of the speech for the differentiation of parkinsonian and control speakers but confirm previous findings about the relevance of plosives and fricatives in the detection of parkinsonian dysarthria.
Analysis of video-oculographic registers for the discrimination of Parkinson’s disease

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, and Juan Ignacio Godino-Llorente

In School and symposium on advanced neurorehabilitation (SSNR), Jul 2018

2017

On the design of a voice pathology assessment system based on the GRB scale.

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, Janaína Mendes-Laureano, and Juan Ignacio Godino-Llorente

In Proceedings and Report - 10th International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications, MAVEBA 2017, Jul 2017

Abs

The purpose of this paper is to present some preliminary results of an automatic system capable of modelling the perceptual abilities of a speech therapist that describes vocal quality in accordance to the GRB scale. The system is trained using three databases that have been evaluated by the same evaluator. 11 spectral, cepstral, modulation spectra and perturbation features are extracted from the input speech. Filter ranking algorithms are utilized to select the most consistent set of features among databases. Decision making is carried out using ordinal classification, to account for the ordering in the GRB scale, and Gaussian regression, to provide continuous decisions about voice quality. Results indicate that the proposed system is proficient when modelling, either by means of the Gaussian regressor and the ordinal classifier, the perceptual abilities of the evaluator. On average the deviations from the actual and predicted label are about half an unit.
Use of acoustic landmarks and GMM-UBM blend in the automatic detection of Parkinson’s disease

Laureano Moro-Velázquez, Juan Ignacio Godino-Llorente, Jorge Andrés Gómez-García, Jesús Villalba, Stephanie Shattuck-Hufnagel, and Najim Dehak

In Proceedings and Report - 10th International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications, MAVEBA 2017, Dec 2017

Abs DOI

New tools based on speech analysis can improve and accelerate diagnosis of Parkinson’s Disease. In this work, the use some specific segments of speech, around the so called Acoustic Landmarks, are used with different families of features such as acoustic cues or Rasta-PLP and GMM-UBM-Blend classification methods to detect Parkinson’s Disease. Results of 87% of accuracy are obtained. Burst segments provide the most relevant information when detecting Parkinson’s Disease while GMM-UBM-Blend is revealed as a promising technique when using small databases and segmented speech.

2016

Detection of Parkinson’s disease by means of GMM-UBM and iVectors techniques

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, Rafael Orozco-Arroyave, and Juan Ignacio Godino-Llorente

In XXXI Simposium Nacional de la Unión Científica Internacional de Radio, URSI 2016, Dec 2016

2015

Tuning of modulation spectrum parameters for voice pathology detection

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, and Juan Ignacio Godino-Llorente

In Models and Analysis of Vocal Emissions for Biomedical Applications: 9th International Workshop, Dec 2015

Abs

The Models and Analysis of Vocal Emissions with Biomedical Applications (MAVEBA) workshop came into being in 1999 from the particularly felt need of sharing know-how, objectives and results between areas that until then seemed quite distinct such as bioengineering, medicine and singing. MAVEBA deals with all aspects concerning the study of the human voice with applications ranging from the neonate to the adult and elderly. Over the years the initial issues have grown and spread also in other aspects of research such as occupational voice disorders, neurology, rehabilitation, image and video analysis. MAVEBA takes place every two years always in Firenze, Italy.
Automatic age detection in normal and pathological voice

Jorge Andrés Gómez-García, Laureano Moro-Velázquez, Juan Ignacio Godino-Llorente, and Germán Castellanos-Domínguez

In 6th Annual Conference of the International Speech Communication Association (INTERSPEECH 2015), Dec 2015

Abs

Systems that automatically detect voice pathologies are usually trained with recordings belonging to population of all ages. However such an approach might be inadequate because of the acoustic variations in the voice caused by the natural aging process. In top of that, elder voices present some perturbations in quality similar to those related to voice disorders, which make the detection of pathologies more troublesome. With this in mind, the study of methodologies which automatically incorporate information about speakers’ age, aiming at a simplification in the detection of voice disorders is of interest. In this respect, the present paper introduces an age detector trained with normal and pathological voice, constituting a first step towards the study of age-dependent pathology detectors. The proposed system employs sustained vowels of the Saarbriicken database from which two age groups are examinated: adults and elders. Mel frequency cepstral coefficients for characterization, and Gaussian mixture models for classification are utilized. In addition, fusion of vowels at score level is considered to improve detection performance. Results suggest that age might be effectively recognized using normal and pathological voices when using sustained vowels as acoustical material, opening up possibilities for the design of automatic age-dependent voice pathology detection systems.

2014

Analysis of complexity and modulation spectra parameterizations to characterize voice roughness

Laureano Moro-Velázquez, Jorge Andrés Gómez-García, and Juan Ignacio Godino-Llorente

In IberSpeech 2014, Dec 2014

2010

Evaluation of feature extraction techniques on event-related potentials for detection of attention-deficit/hyperactivity disorder

Paola Castro-Cabrera, Jorge Andrés Gómez-García, Francia Restrepo, Oscar Moscoso, and Germán Castellanos-Domínguez

In 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology, Aug 2010

Abs DOI

Event-related potentials (ERPs) are one of the most informative and dynamic methods of monitoring cognitive processes, which are widely used in clinical research to deal a variety of psychiatric and neurological disorders as attention-deficit/hyperactivity disorder (ADHD). This work proposes an extraction and selection methodology for discriminating between normal and pathological patients with ADHD by using ERPs. Three different sets of features (morphological, wavelets, and nonlinear based) are analyzed, looking for the best classification accuracy. The results show that the wavelet features provided a good discriminative capability, but it improved by combining all the set of features and applying a feature selection algorithm, reaching a maximum accuracy rate of 91.3%.