Comparative analysis of melodic pattern matching techniques for performance evaluation of singer

Music has always been the vital part of entertainment. Musical schools and singing reality shows are very popular and have become an integral part of our daily life. They provide entertainment to the viewers and a platform to the willing contestants to show their talent. This paper introduces the idea for evaluating the performance of singers using melodic pattern matching techniques, namely LBDM (Local Boundary Detection Model) and DTW (Dynamic Time Warping) based on peaks extracted from the strength and pitches of the audio signal respectively. The audio corpus both standard and recorded song track from singer is maintained using SQL database consist of 5 different song track in MP3 format with duration of approximately 2.10 minutes. The experimental analysis provide the following observation from the comparison between LBDM and DTW with similar and different track singer in term of percentage efficiency; LBDM shows 99.99% matching and DTW shows 99.16% matching, whereas LBDM shows 45.89% matching and DTW shows 55.8% matching with similar and different track singers respectively. The experimental results clearly evident that LBDM provides significant results as compared to DTW.