Method and system device for retrieving songs based on voice modes

Abstract

The invention provides a method and system device for retrieving music based on voice modes. The invention aims at designing a method for retrieving music and songs, which can realize the interaction with a computer based on voice and ensure that the computer can actively recognize inflection information of the voice. In the invention, the technology is also implemented on the computer so as to generate a music retrieval system which can be used for KTV song selection, construction of entertainment websites and mobile terminals. The music retrieval system mainly comprises an interaction interface module, a background processing flow module, a music feature library creation module and a transmission channel module. A user can sing on site after clicking a button; the system can record the voice input in real time, can save a record file and process the file after the recording process is finished, and can ultimately sequence names of songs according to similarity; and one song can be played after being clicked, and relevant information of the song can be displayed. If the first retrieval fails, an additional retrieval can be performed, i.e. a cumulative retrieval can be performed on the basis of the previous retrieval by additionally humming/singing another rhythm of the song.

Description

technical field [0001] The invention relates to the field of file search, in particular to a song search method and system device. Background technique [0002] A search engine is a commonly used resource acquisition method. In the field of digital multimedia, at present, to obtain a specific resource, a key character related to the resource is used for retrieval. With the research on content-based retrieval methods in recent years, the application in the field of audio recognition and retrieval has only begun. Content-based audio recognition and retrieval technology is an important technology for processing massive audio data. Juxtaposition has become a hotspot in content-based retrieval research. Combining audio recognition and retrieval technology with traditional text retrieval can greatly improve the efficiency and accuracy of audio data retrieval and reduce retrieval costs. Content-based music retrieval has also become an emerging research field. With the emergence o...

AI Technical Summary

Problems solved by technology

[0007] 1). At present, although there are multiple test systems, they all have certain limitations. For example, users need to hum fixed symbols, or add auxiliary means, which are difficult for users to use;

[0008] 2). There is still a lot of room for improvement in feature selection. The pitch extraction lacks post-processing, and the pitch sequence cannot directly participate in the matching;

[0009] 3). In the search algorithm, music retrieval mostly uses the traditional string approximate matching algorithm, which does not take into account the characteristics of the music characteristics. Small parts are similar;

[0010] 4). The matching in the retrieval process is not adjusted to be carried out according to short sentences. People generally hum in short sentences, and general retrieval methods do not consider this characteristic of music;

[0011] 5). There is a lack of a standard music database or a method for creating a standard music feature library

At present, there is no universal database that saves the features and lyrics of music scores in units of sentences

Images