Speaker, noise, and acoustic space adaptation for emotion recognition in the automotive environment

Emotional surveillance of drivers possesses significant potential for increased security within passenger transport. In an automotive setting the interaction can further be improved by social awareness of an MMI. Within this scope the detection of security relevant behavior patterns as aggressiveness or sadness is discussed. The focus lies on real-life usage respecting online processing, subject independency, and noise robustness. The approach introduced employs multivariate time-series analysis by brute-force feature generation. Extensive results are reported on two public standard corpora. The influence of noise is discussed by representative car-noise overlay. Thereby impact per low-level-descriptor is considered.