Abstract:
This project deals with the study of Bangla vowel analysis and synthesis which is the basis of Bangla speech processing. The main task is to acquire the Bangla vowel samplesthen study the process of analysis and synthesis using Wavelet Transform (WT). This thesis will explore theory of WT, how to apply WT for Bangla vowel analysis and synthesis, measure the performance of WT for reconstructing the vowels. The performance of WT for synthesizing vowel was measured by calculating Normalized Root Mean Square (NRMSE) between the original and synthesized signal. It is observed from our study that WT with Daubechies 4 (db4) wavelet at decomposition level 5 reproduces the signal with a very small NRMSE in the order of 10-11. Therefore we may say that WT is an appropriate method for Bangla vowel analysis and synthesis.