Infant cry recognition based on feature extraction

Abstract

The choice of feature parameters is very crucial for the system of infant cry recognition. This paper describes the principle and method of mel cepstrum coefficients which is used commonly and adopts a combination algorithm of mel frequency cepstrum coefficients, cepstrum mean subtraction, inverted mel frequency cepstral coefficient, which aim to enhance recognition accuracy in the infant cry recognition system. Results show that the recognition rate of improved algorithm increases compared to the classical algorithm in the same environment.