Turkish emotional speech database

Success of an emotion recognition systems from speech signal is directly dependent to the database used in system modeling as in any pattern recognition problem. In this work, we give the detailed description of the new Turkish emotional speech database we created. Our database consists of 5304 speech signal and their textual contents extracted from 55 Turkish movies. Speech signals are labeled by numerous evaluators both categorically (happy, surprised, sad, angry, fear, neutral and other) and in 3-dimensional emotional space (valence, activation and dominance). We believe that our database will be very helpful in further studies on acoustical analysis of Turkish emotional speech and emotion recognition.