Anti-noise Power Normalized Cepstral Coefficients for Robust Environmental Sounds Recognition in Real Noisy Conditions

This paper proposes a new robust environmental sounds recognition technology based on APNCC to improve the accuracy of environmental sounds recognition in real noisy conditions. First, a highly non-stationary noise estimation algorithm is applied for the noise power spectrum estimation. Second, to achieve noise reduction with less residual colored noise, we present a multi-band spectral subtraction. Then, the process of PNCC extraction is combined with the estimated clean environmental sounds to extract APNCC. Finally, using 70 subclasses of 4 classes of clean environmental sounds, the comparison experiments in different environments under different SNRs are constructed based on the combination of SVM classifier and different features, namely APNCC, PNCC and MFCC. The experimental results show that APNCC outperforms other features in average recognition accuracy and noise robustness, especially for conditions of SNRs lower than 30dB.