Audio-based Speech Emotion and Stress Recognition
Speech, Body and Gesture Languages are familiar methods for inter-human communications. Speech signals carry linguistic information as well as paralinguistic details about the speaker’s emotional state. Speech emotion recognition is a sophisticated method to detect psychological states non-invasively in various contexts. Besides textual content, speech communicates different indicators for human emotions which have a great role in perception and interpretation. The same content may have quite different meanings if it is spoken with different emotional
articulations. The goal of audio-based state recognition is for example to enhance the human-computer interaction between the user and advanced driver assistance systems by developing a speech emotion classifier to evaluate the affective content in speech interaction. Besides emotional states, cognitive workload, driver vigilance and stress can be extracted from speech analysis by comparing rhythm, timbre, melody, tempo, loudness and articulation to pre-classified models.