Speech waveform reconstruction from speech parameters for an effective text to speech synthesis system using minimum phase harmonic sinusoidal model for Punjabi

Speech processing plays a vital role in current speech communication applications. The major objective of digital speech is transmission of messages among human and computer systems. A Text-to-speech synthesizer is utilized for these transmission of speech. Many significant works are carried out in the previous speech synthesis framework but still having issues in quality speech generation. This work presents an effective text to speech synthesis system using minimum phase harmonic sinusoidal modeling in Punjabi language. To develop the presented system, initially input phoneme speech is converted into an overlapping frames. Then, spectral subtraction technique is utilized to eliminate the noise of overlapped frames for speech enhancement. Subsequently, spectral speech parameters such as Mel frequency cepstral coefficients (MFCC), excitation parmeters such as fundamental frequency, energy and their first order(delta) and second order (delta-delta) time derivatives are extracted. After extracting all the parameters, the speech signal is synthesized from the extracted speech parameters using minimum phase harmonic sinusoidal model. Here, this synthesis model is dependent on the extracted speech parameters and also the amplitudes, phases of sine waves for the production of synthetic speech.The presented speech waveform reconstruction method to speeh synthesis process is implemented in PYTHON platform. The experimental outcomes of the presented methodology proved that the presented work is significantly better in terms of various effective performance measures like execution time (0.125 s), SNR (16.4 dB), RMSE (0.074 dB), BSD (3.21%), accuracy (97.6%).