Gammatonegram based speaker identification
Aref Farhadipour, Mohammad Asgari, Mohammad Reza Hasanabadi · 2014
Speech signals contains important information to use for different purposes such as surveillance, smart home, medicine, etc. Thus, classification of these signals are chief to consider for further applications. This article presents a simple method that has lower calculations for center process unit to achieve results as well as faster reaction time and high accuracy. Many method exist in speech signal classification which have their own advantages and disadvantages with regard to their application. In this paper, we propose a new method in speech classification based on human auditory system mixed to spectrogram features. The basilar membrane filters or auditory filters are often modeled by a gammatone function which provides a good approximation to experimentally determined responses. The filter bank derived from these filters is referred as a gammatone filterbank and since STFT can also be likened to a filterbank analysis, these will be an interesting link between standard STFT and gammatone filterbank. A popular tool in time-frequency analysis is spectrogram. Using gammatone filterbank in spectrogram, the method interchanges the fixed filters in all bands and approaches a new diagram called Gammatonegram.