Robust Emotional Stressed Speech Detection Using Weighted Frequency Subbands

Abstract

The problem of detecting psychological stress from speech is challenging due to differences in how speakers convey stress. Changes in speech production due to speaker state are not linearly dependent on changes in stress. Research is further complicated by the existence of different stress types and the lack of metrics capable of discriminating stress levels. This study addresses the problem of automatic detection of speech under stress using a previously developed feature extraction scheme based on the Teager Energy Operator (TEO). To improve detection performance a (i) selected sub-band frequency partitioned weighting scheme and (ii) weighting scheme for all frequency bands are proposed. Using the traditional TEO-based feature vector with a closed-speaker Hidden Markov Model-trained stressed speech classifier, error rates of 22.5/13.0% for stress/neutral speech are obtained. With the new weighted sub-band detection scheme, closed-speaker error rates are reduced to 4.7/4.6% for stress/neutral detection, with a relative error reduction of 79.1/64.6%, respectively. For the open-speaker case, stress/neutral speech detection error rates of 69.7/16.2% using traditional features are used to 13.1/4.0% (a relative 81.3/75.4% reduction) with the proposed automatic frequency sub-band weighting scheme. Finally, issues related to speaker dependent/independent scenarios, vowel duration, and mismatched vowel type on stress detection performance are discussed.

Physical task stress and speaker variability in voice quality

“…An additional application domain included advancements in automatic detection of speech under stress using signal processing advancements derived from the Teager energy operator (TEO) [59], TEO-CB-AutoEnv ( [60]). More recently, nonlinear TEO-based advancements have been considered for stress detection using sub-band filterbank weighting for various actual speech under stress scenarios [61,62]. While these have explored a range of stress conditions, specific speech under physical task stress was not addressed.…”

How to categorize emotional speech signals with respect to the speaker's degree of emotional intensity

“…In addition to these efforts, in order to prepare appropriate systems for being used in more real phenomena, some researchers worked on the topic of ESR in the presence of background noise [10][11][12][13]. The main purpose of these efforts was related to the robustness-improvement of ESR systems.…”

Contribución al estudio de selección de parámetros para identificación de estrés en la voz

“…Algunos ejemplos presentados son grabaciones durante una entrevista de trabajo a diez mujeres y cuatro hombres [106]. Otro ejemplo consistiría en realizar registros durante un examen oral en frente de un jurado [70], el discurso es grabado junto a otras métricas tales como tasa cardíaca, presión sanguínea y tono de voz. Los sujetos participantes fueron seis soldados evaluados por un jurado militar, donde dichos jueces eran de mayor nivel jerárquico que los sujetos en estudio.…”

“…En [70], la detección automática de estrés en la voz se estudia utilizando TEO (Teager Energy Operator ) como característica de extracción y modelos ocultos de Markov como algoritmo de clasificación. Las cadenas ocultas de Markov se basan en un proceso doblemente estocástico, donde se asume que sus estados son desconocidos, los cuales siguen un proceso de Markov.…”

“…Asociado a cada estado se genera un proceso aleatorio el cual crea una secuencia de observación temporal de símbolos [147]. En [70] se consiguió una tasa de error del 13.1 % y de un 4 % en el estado neutro en el mejor de los casos. Pero como es muy común en el estudio del estrés, el corpus utilizado era muy limitado, contando tan solo con seis muestras de locutores, por lo tanto no recaba suficiente validez estadística por el número de sujetos utilizados.…”