Automatic speech segmentation to improve speech synthesis performance

Emerging growth of information and communication technologies has influenced the research trends to focus on speech technologies. Pre-processing of speech signal serves various purposes in any speech processing application. It includes noise removal, endpoint detection, pre-emphasis, framing, windowing, echo canceling etc. Out of these, automatic word/sentence boundary detection is the fundamental step for applications like speech recognition and speech synthesis. This paper expose the problem of automatic words and sentences boundary detection in the silent and noisy situations. This study proposes an algorithm for automatic segmentation of Indian languages voiced speech. A modified data based scheme of finding the entropy of the speech data is placed for better performance. The entropy based method expose good performance features than the energy-based methods. To determine the candidates speech segments, adaptive threshold is used which are related to words and sentences. Simulation results expose that this algorithm will provide better performance than energy based algorithms.