A novel speaker identification system using feed forward neural networks

摘要

This paper proposes a novel speaker identification system which uses Mel Frequency Cepstral Coefficients (MFCC) and Feed Forward Neural Networks (FFNN) for feature extraction and speaker classification respectively. Fuzzy C Mean Clustering (FCM) method is also used against the extracted features from the speech, which facilitates in grouping large amount of data. The efficiency of the system is enhanced furthermore by identifying the gender of the speaker, before the actual speaker identification process, using another FFNN. As a result, the system shows better performance in terms of computational cost and real time identification.