A Cross-language Music Retrieval Method by Using Misheard Lyrics

With the explosion of music data. Traditional music retrieval methods can no longer meet people’s needs. The current music retrieval method is mainly based on the title of the music, the singer, or based on the waveform of the music listened to. But at present, there are often cases of searching for music only by misheard lyrics, especially cross language searching. This greatly increases the difficulty of music retrieval. Therefore, this article first proposes a TextToHumming framework, through which this article converts the misheard lyrics into corresponding music humming records. Then a floating coefficient is added to the DTW algorithm, and the database is retrieved according to the characteristics of the audio. So as to solve this problem. In our datasets, the music retrieval accuracy rate of Top-10 Music reached 88%.