Music humming searching method conducting matching based on binary approach dynamic time warping
A dynamic time normalization and bisection approximation technology, applied in speech analysis, speech recognition, special data processing applications, etc., can solve the problem of insufficient matching speed, unable to solve the problem of relative pitch of notes, and achieve the effect of solving the problem of relative pitch of notes
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Publication Date
- 2014-02-05
- Estimated Expiration
- Not applicable · inactive patent
Smart Images
Figure 1 Figure 2 Figure 3
Abstract
Description
technical field
[0001] The invention relates to a music humming retrieval method based on binary approximation dynamic time rounding matching, and belongs to the technical field of audio retrieval and humming retrieval. Background technique
[0002] The traditional method of retrieving music is text-based retrieval. The current mainstream search engines, such as Baidu, Yahoo, and Google, are very powerful in the field of text retrieval, but the method of retrieving multimedia data is based on text retrieval. It is an inevitable trend in the development of information technology to study more efficient multimedia information retrieval technology based on human communication habits. As one of the important components of multimedia information retrieval, audio retrieval is an important topic in the field of information retrieval technology at home and abroad.
[0003] Humming retrieval is a branch of audio retrieval. When a user retrieves a piece of music using a search engin...
Examples
Embodiment 1
[0071] The present invention proposes a music retrieval method based on binary approximation and dynamic time rounding. The method mainly includes two parts. The first part is to construct a music database through MIDI music files. The second part is to extract the features of the humming melody segment, and perform matching with the music database model based on binary approximation dynamic time normalization and return the retrieval results. This part mainly includes the following steps: firstly, normalize the humming melody segment , audio denoising, pre-emphasis, windowing and framing and other processing to obtain basic features; then perform operations such as filtering out silent segments, pitch detection, and median smoothing to obtain advanced features of humming melody segments; finally humming melody The advanced features extracted from the clips are matched with the template features of the music database one by one based on binary approximation dynamic time roundin...