Journal of Harmonized Research in Engineering

A SENTENCE-PITCH-CONTOUR MODEL FOR BODO LANGUAGE USING HIDDEN MARKOV MODEL (HMM) AND VECTOR QUANTIZATION (VQ)

Abstract

Author(s): Laba kr. Thakuria

Through this paper weintroduce a concept to implement a sentence's pitch-contour model with sentence-wide optimization. This is also called the sentence pitch-contour using HMM(Hidden Markov Model) & VQ (vector quantization) . Here each training sentence are normalized for the pitch-contours of the syllables. This is basically effective for pitch height normalization and after normalization, the pitch-contour of each training syllable is then vector quantized(VQ). The quantization code and lexical tones of adjacent syllables arethen combined to define for HMM training. Using a dynamic-programming in the synthesis phase, the probable observation sequence is produced by finding the sentence wide largest probability path.The pitch-contours of the syllables comprising a sentence which play the main dominant role for the naturalness of the synthesized speech.