Classification and Separation Techniques based on Fundamental Frequency for Speech Enhancement
Archivos
Fecha
2016-01-11
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Jaén : Universidad de Jaén
Resumen
[ES] En esta tesis se desarrollan nuevos algoritmos de clasificación y mejora de voz basados en las propiedades de la frecuencia fundamental (F0) de la señal vocal. Estas propiedades permiten su discriminación respecto al resto de señales de la escena acústica, ya sea mediante la definición de características (para clasificación) o la definición de modelos de señal (para separación).
Tres contribuciones se aportan en esta tesis: 1) un algoritmo de clasificación de entorno acústico basado en F0 para audífonos digitales, capaz de clasificar la señal en las clases voz y no-voz; 2) un algoritmo de detección de voz sonora basado en la aperiodicidad, capaz de funcionar en ruido no estacionario y con aplicación a mejora de voz; 3) un algoritmo de separación de voz y ruido basado en descomposición NMF, donde el ruido se modela de una forma genérica mediante restricciones matemáticas.
[EN]This thesis is focused on the development of new classification and speech enhancement algorithms based, explicitly or implicitly, on the fundamental frequency (F0). The F0 of speech has a number of properties that enable speech discrimination from the remaining signals in the acoustic scene, either by defining F0-based signal features (for classification) or F0-based signal models (for separation). Three main contributions are included in this work: 1) an acoustic environment classification algorithm for hearing aids based on F0 to classify the input signal into speech and nonspeech classes; 2) a frame-by-frame basis voiced speech detection algorithm based on the aperiodicity measure, able to work under non-stationary noise and applicable to speech enhancement; 3) a speech denoising algorithm based on a regularized NMF decomposition, in which the background noise is described in a generic way with mathematical constraints.
[EN]This thesis is focused on the development of new classification and speech enhancement algorithms based, explicitly or implicitly, on the fundamental frequency (F0). The F0 of speech has a number of properties that enable speech discrimination from the remaining signals in the acoustic scene, either by defining F0-based signal features (for classification) or F0-based signal models (for separation). Three main contributions are included in this work: 1) an acoustic environment classification algorithm for hearing aids based on F0 to classify the input signal into speech and nonspeech classes; 2) a frame-by-frame basis voiced speech detection algorithm based on the aperiodicity measure, able to work under non-stationary noise and applicable to speech enhancement; 3) a speech denoising algorithm based on a regularized NMF decomposition, in which the background noise is described in a generic way with mathematical constraints.
Descripción
Palabras clave
Clasificación de voz, Mejora de voz, Frecuencia fundamental, Factorización de matrices no negativas, Audífonos digitales, Speech classification, Speech enhancement, Fundamental frequency, Nonnegative matrix factorization (NMF), Hearing aids
Citación
Cabañas-Molero, Pablo-Antonio. Classification and Separation Techniques based on Fundamental Frequency for Speech Enhancement. 2016, 196 p. [http://hdl.handle.net/10953/740]