Speaker Identification in Emotional Environments

The performance of speaker identification is almost perfect in the neutral environment. However, the performance is significantly deteriorated in emotional environments. In this work, three different and separate models have been used, tested and compared to identify speakers in each of the neutral and emotional environments (completely two separate environments). Our emotional environments in this work consist of five emotions. These emotions are: angry, sad, happy, disgust and fear. The three models are: Hidden Markov Models (HMMs), Second-Order Circular Hidden Markov Models (CHMM2s) and Suprasegmental Hidden Markov Models (SPHMMs). Our results show that the three models perform extremely well for speaker identification in the neutral environment. In emotional environments, the average speaker identification performance based on HMMs, CHMM2s and SPHMMs is 61.4%, 66.4% and 69.1%, respectively. Our results in this work are better than those obtained in subjective evaluation by human judges.