In the Discrete Cosine Transform (DCT) domain, the tones of musical instruments can be divided into two groups. The first one with the single significant local peaks and the second one with the multiple significant local peaks. The second one can be divided into two sub groups, which have many and a few significant local peaks. This research deal with multiple significant local peaks. In this research, segment averaging was used to reduce the number of DCT coefficients, in the DCT domain. In

more »

... s case, the reduced number of DCT coefficients called feature extraction coefficients. Based on the experiment, when the segment averaging of DCT coefficients was used optimally for the tones which had many (i.e. thirteen) and a few (i.e. three) significant local peaks, it could give 8 and 16 feature extraction coefficients respectively. So, in order that segment averaging of DCT coefficients could be used optimally, either for the tones which have many or a few significant local peaks in the DCT domain, it could use segment length 4 points and DCT length 64 points. By using it, it could give 16 feature extraction coefficients.