AM-FM Based Robust Speaker Identification in Babble Noise

Speech babble is one of the most challenging noise interference due to its speaker/speech like characteristics for speech and speaker recognition systems. Performance of such systems strongly degrades in the presence of background noise, like the babble noise. Existing techniques solve this problem by additional processing of speech signal to remove noise. In contrast to existing works, the aim is to improve noise robustness focusing on the features only. To derive robust features, amplitude modulation frequency modulation (AMFM) based speaker model is proposed. The robust features are derived by fusing the characteristics of speech production and speech perception mechanisms. The performance is evaluated using clean speech corpus from TIMIT database combined with babble noise from the NOISEX-92 database. Experimental results show that the proposed features significantly improve the performance over the conventional Mel frequency cepstral coefficient (MFCC) features under mismatched training and testing environments. General Terms Pattern Recognition, Algorithms, Experimentation.