Performance Comparison Of 2-D DCT On Full/Block Spectrogram And 1-D DCT On Row Mean Of Spectrogram For Speaker Identification

The goal of this paper is to present a very simple approach to text dependentspeaker identification using a combination of spectrograms and well knownDiscrete Cosine Transform (DCT). This approach is based on use of DCT tofind similarities between spectrograms obtained from speech samples. Theset of spectrograms forms the database for our experiments rather than rawspeech samples. Performance of this approach is compared for differentnumber of coefficients of DCT when DCT is applied on entire spectrogram,when DCT is applied to spectrogram divided into blocks and when DCT isapplied to the Row Mean of a spectrogram. Performance comparison showsthat, number of mathematical computations required for DCT on Row Mean ofspectrogram method is drastically less as compared to other two methodswith almost equal identification rate.