SUK · Third Year B.Tech. Computer Science and Engineering (Artificial Intelligence and Machine Learning) · NEP 2020
Speech Systems
AIML3205 · Elective III · Semester VI · 3 credits · ISE 10 + mid-semester 30 + ESE 60
One of the Elective III options.
Unit 1: Introduction to Speech Systems
6 hoursSpeech production and its discrete-time model; perception and the auditory system; articulatory, acoustic and auditory phonetics; speech sounds and spectrograms; pitch measurement in the spectral and cepstral domains; formants of voiced and unvoiced speech.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Articulatory and acoustic phonetics
- Formants
- Pitch period estimation by autocorrelation and cepstrum
Unit 2: Speech Signal Analysis and Feature Extraction
5 hoursShort-time analysis, windowing, energy and zero-crossing rate; the STFT, spectrograms and filter banks; LPC, cepstral analysis and pattern comparison.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Short-time energy and zero-crossing rate
- Short-time Fourier transform, filter banks and LPC
- LPC, PLP and MFCC features
- Speech distortion measures and cepstral distances
Unit 3: Speech Modelling
6 hoursMarkov processes; HMM components, training, Viterbi search and Baum-Welch re-estimation; GMMs, SVMs and neural approaches; vector quantisation and codebook design with k-means and LBG.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- HMMs for speech: evaluation and Viterbi decoding
- Baum-Welch re-estimation
- Gaussian mixture models
- Learning vector quantisation
Unit 4: Speech Recognition Systems
6 hoursASR architecture with acoustic and language models; isolated, connected and large-vocabulary continuous recognition; n-grams and context-dependent units; end-to-end deep-learning ASR.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Automatic speech recognition
- N-grams, smoothing, interpolation and backoff
- Dynamic time warping
Unit 5: Speech Enhancement
7 hoursClasses of enhancement algorithms; multiband, MMSE and perceptual spectral subtraction; Wiener filters in time and frequency; maximum-likelihood, Bayesian, MMSE and log-MMSE estimators; subspace algorithms.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Audio enhancement: spectral subtraction and Wiener filtering
Unit 6: Speech Synthesis and Applications
6 hoursText-to-speech systems; concatenative synthesis using formants and LPC; applications.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Text-to-speech synthesis
Unit contents are summarised from the official SUK syllabus . AiCanCode.org is not affiliated with Shivaji University, Kolhapur.