An Approach of Enhanced PNCC for Resident Identification Applications

The performance of applying voice control in home automation can significantly drop under multi-resident situations and noisy environments. It therefore requires some appropriate approaches for smart home applications to address the problem of resident identification. Voice recognition, which explores characteristics of voice, is a potential biometric modality for such problem in smart home. In this paper, the power-normalized cepstral coefficient (PNCC) of voice biometrics is applied to identify individuals in smart home. A new technique of power-law nonlinearity and an algorithm of noise suppression based on asymmetric filtering are used to enhance feature extraction and reduce environmental noise. This proposed approach extremely reduces error rate and achieves high performance on different data sets.