Audio retrieval based on wavelet transform
Deqin Chen, Wenhui Zhang, Zhibo Zhang, Wei Huang, Jia Ao · 2017
The focus of this paper is how to use a smaller audio to retrieve a correct result. It presents an audio retrieval algorithm based on wavelet transform that can shrink the duration of audio retrieval to 1s while ensure a high accuracy. In addition, this algorithm which analyses time-frequency based on wavelet transform can not only carry out the correct retrieval results, but also accurately locate the correct position in the results. The experiments set up an audio library consisted of 200 WAV format music. Then the audio clip with the length of 1s, 2s, 3s, 4s and 5s is randomly selected to test the performance of the algorithm via format covert, adding 20dB, 30dB, or 40dB white Gaussian noise. Experiments show that the recall rate is 100%, meanwhile the recall rates of our method are all over 91.5% for audios longer than 3 seconds.