Abstract:In the present study, an experiment was conducted to generate F0 contours for Mandarin with a pitch target approximation model proposed in Xu, Xu and Luo (1999). In this model, F0 contours in speech are assumed to be resulting from asymptotic approximation to underlying pitch targets that are either static or dynamic. The model parameters were estimated through nonlinear regression using the Levenberg–Marquardt algorithm. The speech corpus consisted of sentences from Voice of America broadcasting news. After the regression analysis, sentences were re-synthesized with the generated F0 using the TD-PSOLA technique. Preliminary results indicate that F0 contours generated by the model are close to the original both numerically and perceptually. Furthermore, most underlying pitch targets obtained through the regression analysis seem to match the models basic assumptions. However, it is also apparent that information about both higher-level linguistic functionality and additional low-level articulatory constraints is needed to account for the numerical variations in the estimated parameters. In general, the results are encouraging as they show that the model can generate close-fitting F0 contours even with strong linguistic assumptions. This suggests that it has the potential to evolve into a system with the predictive power desirable for intonation modeling.

Generation of Fundamental Frequency Contours for Mandarin Speech Synthesis Based on Tone Nucleus Model

On Fundamental Frequency Contour Synthesis And Control Method For Chinese Speech Synthesis

A Method for Automatic Tone Command Parameter Extraction for the Model of F0 Contour Generation for Mandarin

A Method for Automatic Tone Command Parameter Extraction for the Model of F 0 Contour Generation for Mandarin

Rule-based Generation of Phrase Components in Two-step Synthesis of Fundamental Frequency Contours of Mandarin

Experiment on pitch target approximation model for generating Mandarin F0 contour

Tone nucleus modeling for Chinese lexical tone recognition

A PITCH TARGET APPROXIMATION MODEL FOR F0 CONTOURS IN MANDARIN

Automatic Extraction Of Tone Command Parameters For The Model Of F(0) Contour Generation For Standard Chinese

Effects of fundamental frequency contour on understanding Mandarin sentences in bimodal hearing simulations

Tone Generation by Maximizing Joint Likelihood of Syllabic HMMs for Mandarin Speech Synthesis

Towards the Automatic Extraction of Fujisaki Model Parameters for Mandarin

Identification and Synthesis of Cantonese Tones Based on the Command-Response Model for F0 Contour Generation.

The command-response model for the generation of F0 contours of Cantonese Utterances

Modeling Tone and Intonation in Mandarin and English As a Process of Target Approximation.

Physiological and Physical Mechanisms for Fundamental Frequency Control in Some Tone Languages and a Command-Response Model for Generation of Their F0 Contours

Improving F0 prediction using bidirectional associative memories and syllable-level F0 features for HMM-based Mandarin speech synthesis

Comparison of perceived prosodic boundaries and global characteristics of voice fundamental frequency contours in mandarin speech

Robust F0 Modeling for Mandarin Speech Recognition in Noise.

Generating Mandarin and Cantonese F0 Contours with Decision Trees and BLSTMs