This study is to increase recognition rate and to reduce processing time in the speaker recognition system. For processing time reduction we used only difference between average reference pattern and speaker's input pattern without using DTW which had...
This study is to increase recognition rate and to reduce processing time in the speaker recognition system. For processing time reduction we used only difference between average reference pattern and speaker's input pattern without using DTW which had been mainly used in pattern matching. For increasing recognition rate existing cepstrum weights are modified. Experimental results with ten speakers showed increase by 1.75%(95.25% from 93.5%) in speaker verification and 100% in speaker identification.