Font Size: a A A

A Study On Pitch Based Beautification System Of Singing Voice

Posted on:2017-03-31Degree:MasterType:Thesis
Country:ChinaCandidate:W ShaFull Text:PDF
GTID:2308330485951004Subject:Control engineering
Abstract/Summary:PDF Full Text Request
Singing voice is rhythmic sound emitted by humans who consciously control the vibration of their vocal cords and the deformation of their nose, mouth and other speech organs. The improper control may cause the singing voice to be out of tune or not pleasant. Therefore, it’s good if we can use music editing techniques to beautify the expected singing voice. In the above context, we combine the music editing techniques and the extraction algorithm of speech features to solve the problems of beautifying the singing voice. The topics addressed here are significant for the theoretical exploration and the practical process of the singing voice beautification.The main work and innovations of this paper are as follows:1. In order to gain the desired music materials, we collected 210 folk songs and popular songs, and invited a professional singer to record the song database for up to 132 minutes. These databases have laid a solid foundation for the construction of the pitch based singing voice beautification system. In order to verify the effectiveness of the singing voice beautification system, we specially invited three students (one with normal intonation, the other two with abnormal intonation) to participate in the recording, each of them recorded the 12 songs as researching materials.2. We studied the beautification problem of the out of tune singing voice. When the music score is known, we have proposed a pitch based singing voice beautification method. The method can generate the singing voice consistent with the music score, and provide the algorithm level support for the subsequent singing voice beautification system.3. When the music score is unknown, we have proposed a target pitch based singing voice beautification method. With the help of HTS singing voice synthesis framework, the recorded voice database is used to train the hidden Markov model and generate the target pitches. The method can generate the singing voice consistent with the target music score, and provide the algorithm level support for the subsequent singing voice beautification system.4. Based on the above singing voice beautification algorithm developed, we have constructed a pitch based singing voice beautification system. The subjective and objective evaluation experiments show that the method proposed in this paper can be used to beautify the singing voice.
Keywords/Search Tags:Singing voice, Singing voice beautification system, Singing voice database, MIDI, Hidden Markov model
PDF Full Text Request
Related items