An Efficient Music Retrieval Using Noise Cancellation

In this paper we present a music retrieval method. The algorithm works in two steps: noise cancellation and music retrieval. In noise cancellation step we apply filtering on psychoacoustic representation of the given music signal to get a clean noise-free signal which can truly represent the music. Each music has a particular and unique beat pattern called main beat. This representative feature from the noise-free signal is extracted which is followed by one matching method at the retrieval stage. Similar types of features are extracted from the music database or training music as well as the test music. For feature extraction, a music signal is primarily segmented into small windows in the time domain. Extracted features from both the training and query music are matched using similarity matching algorithm. 94% retrieval accuracy of our method is quiet satisfactory. Our experimental result shows the feasibility of our approach.