Important RGPV Question
EC-803 (C) Speech Processing
VIII Sem, EC
UNIT 1-Basic Concepts Of Speech Processing
Q.1) Define a speech signal.
Q.2) What do you understand by filter? Explain.
Q.3) Explain the concept of complex cepstrum of speech.
Q.4) Differentiate between convolution and deconvolution of speech.
Q.5) Explain in detail the mechanics of speech production and acoustic phonetics.
Q.6) What is short-time Fourier analysis? Explain the properties of short-time Fourier analysis.
Q.7) What is homomorphic system of convolution? Explain in detail.
Q.8) Explain digital models for speech signals using examples.
Q.9) Explain one application of vocoders in short.
Q.10) What is the role of cepstrum in speech processing?
Q.11) Differentiate between speech and silence.
Q.12) Explain acoustic phonetics with proper example.
Q.13) Explain in detail the mechanics of speech production and acoustic phonetics.
Q.14) Show that the log power spectrum of y is approximately that of x with an additional ripple. Find the parameters of this ripple.
UNIT 2-Speech Analysis
Q.1) What are lossless tube models of speech signal?
Q.2) Write short note on speech spectrogram.
Q.3) Define pitch detection.
Q.4) Define correlation function with an example.
Q.5) Write the basic principle of linear predictive coding of speech.
Q.6) Discuss the frequency domain interpretation of mean squared prediction error of a lossless tube model. Also describe the relations between various speech parameters.
Q.7) What is pitch period estimation using parallel processing? Explain with proper equations.
Q.8) What are the factors which have to be considered in automatic recognition of isolation during speech versus silence discrimination? Elaborate with two examples.
Q.9) Multipulse LPC uses an excitation with several pulses per pitch period. Explain how this can improve LPC quality.
Q.10) Explain pitch detection using Homomorphic processing technique.
Q.11) How do we compute the gain for a given model? Explain.
UNIT 3-Speech Modeling
Q.1) Explain the fundamental concepts of a Markov Process. How are these concepts extended to form a Hidden Markov Model (HMM) for speech modeling?
Q.2) Describe the three basic problems associated with Hidden Markov Models in the context of speech processing.
Q.3) Explain the Baum-Welch re-estimation algorithm. What is its role in training HMMs for speech recognition, and what parameters does it re-estimate?
Q.4) Elaborate on the practical implementation issues of Hidden Markov Models for large vocabulary continuous speech recognition systems.
Q.5) Define the forward and backward probabilities in the context of HMMs. How are these probabilities used in the Baum-Welch algorithm?
UNIT 4-Speech Recognition
Q.1) Show that the recognition of speech results in large error because the quantization error is multiplied by the prediction gain. Show that with closed-loop prediction this does not occur.
UNIT 5-Speech Synthesis
Q.1) What is vocoder and channel vocoder? Explain with proper diagrams.
Q.2) With the help of a block diagram explain homomorphic vocoder containing analyzer and synthesizer.
— Best of Luck for Exam —
