This paper presents a new feature extraction method for speech recognition that is robust to additive background noise. The proposed method is based on the autocorrelation mel frequency cepstral coefficients (AMFCCs) and a variable window. The propose...
This paper presents a new feature extraction method for speech recognition that is robust to additive background noise. The proposed method is based on the autocorrelation mel frequency cepstral coefficients (AMFCCs) and a variable window. The proposed method applies a variable window to higher-lag autocorrelation coefficients, depending on the frame energy and periodicity. The performance of the proposed method is verified using an Aurora-2 task, and the results confirm a significantly improved performance under noisy conditions.