Comparison Between Different Methods for Epoch Extraction from Speech Signal

Speech is produced by exciting the vocal tract system by an excitation source. The excitation is significant at the instants of glottal closure. These instants of significant excitation are called Epochs. Epochs play a significant role in speech processing and is used are used in many applications. There are different methods for extraction of epochs from speech signal. In this paper we have discussed and compared three methods namely Linear Prediction Based (LP Based) Method, Group Delay Based (GD Based) Method and Zero Frequency Filtering (ZFF) Method. LP-Based method and GD-Based method depends on characteristics of both the excitation source and vocal tract system. ZFF method is based on the impulse nature of the excitation signal and is not affected by the characteristics of the vocal tract system. The ZFF method gives more accurate and reliable results compared to the first two methods.