Speech Quality Assessment and Emotions - Effect on the PESQ Metric

This paper analyzes how emotional speech influ-ences the quality assessment process. The objective metrics for speech quality assessment are usually evaluated with neutral speech recordings, but human interaction frequently involves emotions. Those can modify speech signals' acoustic and phonetic characteristics largely, which are essential from the objective speech quality assessment perspective. The proposed analysis was carried out with the English and Slovenian Interface emotional speech database, which includes two neutral and six emotional speaking styles. Speex codec was used to simulate the degradations, which were then assessed with the lTV - T P.862 PESQ objective metric. The results' analysis showed that in both languages surprise, in English fear, and in Slovenian disgust influenc the speech quality assessment significantly.