SPPU · Final Year Electronics & Telecommunication Engineering · 2019 Course
Speech Processing
404184A · Elective III · Semester VII · 3 credits · In-semester 30 + end-semester 70
One of the Elective III options.
Unit 1: Fundamentals of Speech Processing
6 hoursSpeech production and its LTI and time-varying models; phonetics and articulation; vowels, semivowels, nasals, stops, fricatives and the vowel triangle; pitch by autocorrelation and AMDF; formants.
Marks: 15 in the in-semester exam
- Speech production model
- Classification of speech sounds
- Pitch detection by autocorrelation and AMDF
- Formants
Unit 2: Time and Frequency-domain Analysis
7 hoursShort-time energy, magnitude and zero-crossing rate; speech/silence discrimination; short-time autocorrelation and AMDF; spectral centroid, spread, entropy, flux and roll-off; narrowband and wideband spectrograms.
Marks: 15 in the in-semester exam
- Short-time energy and zero-crossing rate
- Speech and silence discrimination
- Short-time autocorrelation
- Spectral audio features
- Spectrograms
Unit 3: Linear Prediction and Cepstral Analysis
6 hoursLinear predictive analysis; forming and solving the LPC equations; cepstral analysis and cepstral coefficients; computing MFCCs.
Marks: 17–18 in the end-semester exam
- Linear predictive coding of speech
- Solving the LPC equations
- Cepstral analysis
- Mel frequency cepstral coefficients
Unit 4: Speech and Audio Coding
6 hoursPCM, companded PCM and DPCM; filter-bank, sub-band and adaptive transform coders; harmonic coding; LPC vocoders; phase and homomorphic vocoders.
Marks: 17–18 in the end-semester exam
- Waveform coders: PCM and DPCM
- Sub-band and transform coders
- Vocoders
Unit 5: Applications of Speech Processing
6 hoursAutomatic speech recognition and its features; deterministic and statistical sequence recognition; speaker identification and verification; text-to-speech by concatenative and waveform synthesis; prosody.
Marks: 17–18 in the end-semester exam
- Automatic speech recognition
- Speaker identification and verification
- Text-to-speech synthesis
Unit 6: Speech Processing with Machine Learning
6 hoursASR and TTS with SVMs, CNNs and RNNs; accuracy, confusion matrix, sensitivity, specificity, ROC and AUC.
Marks: 17–18 in the end-semester exam
- Speech recognition with neural networks
- Evaluation metrics: confusion matrix, ROC and AUC
Unit contents are summarised from the official SPPU syllabus . AiCanCode.org is not affiliated with Savitribai Phule Pune University.