A Comparison of Audio Features of Thai Classical Music Instrument

In the presented study, we found the characteristics of Thai classical music affects the similarity measure of music retrieval. On account of the retrieval to provide higher accuracy requires in what the represented the actual music could be. With regards to our experiments, we found that each feature has the capability to retrieve differently. The Experiments are conducted on 60 original music tracks and 180 music clips of Thai classical music. We employ the Euclidean distance, a simple similarity measure, which to find similarity between original whole music and music clips for each song. The experimental results of comparisons Cepstrum, Mel Frequency Cepstral Coefficient (MFCC) and Chromagram Audio Features of Thai Classical Music Instruments showed that Chromagram can achieve the most features selections to correct the similarity rate of 85%. And cosine distance measure has the efficiency to retrieve rate of 71.66%.