SPEECH TECHNOLOGY

Models of human speech production and perception. Preprocessing of the speech signal. Acoustic parameters. Time domain speech coding. Frequency domain speech coding. Artificial neural networks. Speech recognition systems. Speaker recognition systems. Speech synthesis. Applications.

Objectives

• Mechanisms of speech production and perception • Speech preprocessing • Acoustic parameters • LPC and Cepstrum analysis • Formants • Time domain waveform coding, Frequency domain speech coding and Vocoders • Dynamic Time Warping (DTW) • Hidden Markov Models (HMM) • Artificial Neural Networks • Speech and Speaker Recognition Systems • Speech Synthesis and TTS

Prerequisites

There aren't any.

Syllabus

Modeling speech production: Mechanism of speech production, Speech sounds, Model of speech production. Speech preprocessing: Selection of sampling frequency, Digitization, Short-term processing of speech signals, Frame length selection, Pre-emphasis, Filter-design window functions, Frame rate. Acoustic parameters: Energy, Zero-crossings, Fundamental frequency, Methods for tonality calculation, Spectrograph, Formants, Linear prediction coefficients (LPC), Filter bank, Reflection coefficients, Cepstral coefficients. Speech processing techniques: Matching acoustic models, Dynamic time warping (DTW), Vector quantization, k-means Algorithm, VQ codebook with a mixture of densities, Hidden Markov model s (HMM), Forward–backward algorithm, Viterbi algorithm. Speech recognition systems. Speaker recognition systems. Speech synthesis: Basic principles, Unit size, Unit types, Synthesis methods, systems with limited/unlimited vocabulary. Articulatory synthesis, Formant synthesis, LPC synthesis, Modeling the excitation source, Models of prosody and intonation, LPC Cepstrum, ARMA model parameter estimation. Digital noise reduction techniques. Speech coding: Time domain waveform coding, Frequency domain speech coding, Vocoders.