Implementation of Emotion Detection by Cepstral Analysis and Noise Filtration in MATLAB

Abstract

The main objective of the paper is to determine the emotional state of a person, through his speech. The two most important features that are used to identify the emotional states of a person are i. Pitch and ii. Formant  frequencies. To understand the three emotional states of a person, these two pitch and formant frequencies are first extracted from the speech signal and, later their analysis is carried out. The three emotions that are considered are Anger, Neutrality and Happiness. Simulations are carried out in MATLAB and the TU-Berlin database is used for the analysis. For the extraction of pitch frequency Cepstral analysis method is used and the formant frequencies are estimated to detect happy emotions. The objective of cepstral analysis is to separate the speech into its source and system components without any prior knowledge about source and /
or system. This separation is done due to difference in occupancy of frequency ranges. Following it, noise filtration is implemented to remove inaudible ambient noises.