Singing-voice Synthesis Using ANN Vibrato-parameter Models

Vibrato is an important factor that affects the naturalness level of a synthetic singing voice. Therefore, the analysis and modeling of vibrato parameters are studied in this paper. The vibrato parameters of those syllables segmented from recorded songs are analyzed by using short-time Fourier transform and the method of analytic signal. After the vibrato parameter values for all training syllables are extracted and normalized, they are used to train an artificial neural network (ANN) for each type of vibrato parameter. Then, these ANN models are used to generate the values of vibrato parameters. Next, these parameter values and other music information are used together to control a harmonic-plus-noise model (HNM) to synthesize Mandarin singing voice signals. With the synthetic singing voice, subjective perception tests are conducted. The results show that the singing voice synthesized with the ANN generated vibrato parameters is much increased in the naturalness level. Therefore, the combination of the ANN vibrato models and the HNM signal model is not only feasible for singing voice synthesis but also convenient to provide multiple singing voice timbres.