Antonio Satue-Villar

Nonlinear Analyses and Algorithms for Speech Processing

The COST-277 European Action: An Overview.- The COST-277 European Action: An Overview.- Neuro-fuzzy Logic in Signal Processing for Communications: From Bits to Protocols.- Connected Operators for Signal and Image Processing.- Speaker Recognition.- Exploiting High-Level Information Provided by ALISP in Speaker Recognition.- MLP Internal Representation as Discriminative Features for Improved Speaker Recognition.- Weighting Scores to Improve Speaker-Dependent Threshold Estimation in Text-Dependent Speaker Verification.- Parameter Optimization in a Text-Dependent Cryptographic-Speech-Key Generation Task.- The COST-277 Speech Database.- Speech Analysis.- Childrens Organization of Discourse Structure Through Pausing Means.- F0 and Intensity Distributions of Marsec Speakers: Types of Speaker Prosody.- A Two-Level Drive - Response Model of Non-stationary Speech Signals.- Advanced Methods for Glottal Wave Extraction.- Cepstrum-Based Estimation of the Harmonics-to-Noise Ratio for Synthesized and Human Voice Signals.- Pseudo Cepstral Analysis of Czech Vowels.- Bispectrum Estimators for Voice Activity Detection and Speech Recognition.- Voice Pathologies.- On the Acoustic-to-Electropalatographic Mapping.- Issues in Clinical Applications of Bilateral Multi-step Predictive Analysis of Speech.- Optimal Size of Time Window in Nonlinear Features for Voice Quality Measurement.- Support Vector Machines Applied to the Detection of Voice Disorders.- Synthesis of Disordered Voices.- Voice Pathology Detection by Vocal Cord Biomechanical Parameter Estimation.- Speech Recognition.- A Speech Recognizer Based on Multiclass SVMs with HMM-Guided Segmentation.- Segment Boundaries in Low Latency Phonetic Recognition.- Third-Order Moments of Filtered Speech Signals for Robust Speech Recognition.- New Sub-band Processing Framework Using Non-linear Predictive Models for Speech Feature Extraction.- Noise Robust Automatic Speech Recognition with Adaptive Quantile Based Noise Estimation and Speech Band Emphasizing Filter Bank.- Spotting Multilingual Consonant-Vowel Units of Speech Using Neural Network Models.- Speech Enhancement.- Novel Sub-band Adaptive Systems Incorporating Wiener Filtering for Binaural Speech Enhancement.- Maximum a Posterior Probability and Cumulative Distribution Function Equalization Methods for Speech Spectral Estimation with Application in Noise Suppression Filtering.- Applications.- Modeling Fluctuations of Voiced Excitation for Speech Generation Based on Recursive Volterra Systems.- A Simple, Quasi-linear, Discrete Model of Vocal Fold Dynamics.- Blind Channel Deconvolution of Real World Signals Using Source Separation Techniques.- Method for Real-Time Signal Processing Via Wavelet Transform.

Advances in a Multimodal Approach for Dysphagia Analysis Based on Automatic Voice Analysis

Parkinson’s disease (PD) is the second most frequent neurodegenerative disease with prevalence among general population reaching 0.1–1 %, and an annual incidence between 1.3–2.0/10,000 inhabitants. The most obvious symptoms are movement-related such as tremor, rigidity, slowness of movement and walking difficulties and frequently these are the symptoms that lead to the PD diagnoses but also they could have In this sense voice analysis is a safe, non-invasive, and reliable screening procedure for PD patients with dysphagia, which could detect patients at high risk of clinically significant aspiration. In this paper we will present a part of an ongoing project that will evaluate automatic speech analysis based on linear and non-linear features. These can be reliable predictors/indicators of swallowing and balance impairments in PD. An important advantage of voice analysis is its low intrusiveness and easy implementation in clinical practice. Thus, if a significant correlation between these simple analyses and the gold standard video-fluoroscopic analysis will imply simpler and less stressing diagnostic test for the patients as well as the use of cheaper analysis systems.

A Multimodal Approach for Parkinson Disease Analysis

Parkinson’s disease (PD) is the second most frequent neurodegenerative disease with prevalence among general population reaching 0.1-1 %, and an annual incidence between 1.3-2.0/10000 inhabitants. The mean age at diagnosis of PD is 55 and most patients are between 50 and 80 years old. The most obvious symptoms are movement-related; these include tremor, rigidity, slowness of movement and walking difficulties. Frequently these are the symptoms that lead to the PD diagnoses. Later, thinking and behavioral problems may arise, and other symptoms include cognitive impairment and sensory, sleep and emotional problems. In this paper we will present an ongoing project that will evaluate if voice and handwriting analysis can be reliable predictors/indicators of swallowing and balance impairments in PD. An important advantage of voice and handwritten analysis is its low intrusiveness and easy implementation in clinical practice. Thus, if a significant correlation between these simple analyses and the gold standard video-fluoroscopic analysis will imply simpler and less stressing diagnostic test for the patients as well as the use of cheaper analysis systems.

An application to acquire audio signals with chicoplus hardware

In this paper we present a method to acquire the overall signal (signal of interest plus interferences plus noise) impinging upon an array of 4 microphones. This is achieved by using the “Chico Plus” audio recorder card, by Innovative Integration, a Borland C++ Builder 5 program to develop an application for the user communicating with the card, an array of microphones, and some filtering algorithms implemented with MatLab. We also want to compare the results obtained using these real signals with those ones got with simulated signals described in previous work.