Classification of speech under stress based on features derived from the nonlinear Teager energy operator

Abstract

Studies have shown that the distortion introduced by stress or emotion can severely reduce speech recognition accuracy. Techniques for detecting or assessing the presence of stress could help neutralize stressed speech and improve the robustness of speech recognition systems. Although some acoustic variables derived from linear speech production theory have been investigated as indicators of stress, they are not consistent. Three new features derived from the nonlinear Teager (1990) energy operator (TEO) are investigated for stress assessment and classification. It is believed that TEO based features are better able to reflect the nonlinear airflow structure of speech production under adverse stressful conditions. The proposed features outperform stress classification using traditional pitch by +22.5% for the normalized TEO autocorrelation envelope area feature (TEO-Auto-Env), and by +28.8% for TEO based pitch feature (TEO-Pitch). Overall neutral/stress classification rates are more consistent for TEO based features (TEO-Auto-Env: /spl sigma/=5.15, TEO-pitch: /spl sigma/=7.83) vs. (pitch: /spl sigma/=23.40). Also, evaluation results using actual emergency aircraft cockpit stressed speech from NATO show that TEO-Auto-Env works best for stress assessment.