Determining the voiceprint recognition on the basis of emotional speech signal: Indonesia language

摘要

Automatic voiceprint recognition, posited on human speech signal, serves many salient practical applications. A number of studies are undertaken on the basis of normal speech. This research intends to develop automatic voiceprint recognition system on the basis of emotion speech signal in Indonesia language. The study is limited to four different people with speeches of four distinctive emotional conditions, i.e. happy, sad, angry, and fear. The 48 voiceprint features from emotion-related speeches data are extracted by applying Mel Frequency Cepstral Coefficient, speaker classification utilized these features. The categorization is also bolstered by Support Vector Machine method. The results suggested that the recognition managed to achieve about 92% of the level of accuracy.