Speech synthesis in Indian languages

Abstract: This paper presents the study of phonemes in the Indian languages for developing good quality speech synthesis. Harmonic plus noise model (HNM) which divides the speech signal in two sub bands: harmonic and noise, is implemented with the objective of studying its capabilities and to investigate the adaptation needed. Childers and Hu's algorithms are used for voicing and pitch detection. As the analysis and synthesis of the speech is pitch synchronous, so glottal closure instants (GCIs) should be accurately calculated. For comparison, impedance glottograph is also used to extract the GCIs. Investigations show that HNM is capable of synthesizing all syllables and passages spoken in different styles in Indian languages with natural sounding output. The quality of synthesized speech improves if the GCIs are taken from the glottal signal instead of being obtained by processing the speech signal. As HNM uses amplitudes and phases of the pitch harmonics, interpolating the spectrum at the desired pitch harmonics and synthesizing using these values can achieve pitch scaling