II Semester M.E. (Electronics & Communication) Degree Examination, January 2015 (2K9 Scheme) EL 215 : SPEECH AND AUDIO SIGNAL PROCESSING Time : 3 Hours
Max. Marks : 100
Instruction : Answer any five full questions. 1. a) With a neat block diagram, explain the speech production mechanism in terms of i) excitation along with the type of speech sounds produced and ii) system (filter). b) Write a note on : Articulators.
8 3
c) Explain the following : i) Formants, ii) Pitch, and iii) Spectrograms.
9
2. a) Define short time autocorrelation function. Discuss the clipping autocorrelation pitch detector algorithm.
8
b) Explain the significance of short time energy and short time average zero crossing rate in the context of speech signal processing.
6
c) Show that the short time energy can be expressed recursively as, En = a2 En–1 + x2 (n), where the window h (n), is defined as h (n) = anu(n), x (n) is the speech signal, and 0 < a < 1. Draw a digital network diagram of this equation.
6
PED – 136
-2-
*PED136*
3. a) Explain short time Fourier transform. Discuss the Fourier transform interpretation of STFT, specifically in terms of i) periodicity, ii) & recovering the signal x (n) from its STFT.
6
b) Discuss the filter bank summation method of short time synthesis and derive the expression for necessary constraints.
8
STFT c) Let x(n) ←⎯ ⎯ ⎯→ Xn (e jω)
i) If v(n) = x (n) + y (n), then show that
( )
( ) ( )
Vn e jω = Xn e jω + Yn e jω .
ii) If v (n) = x (n – k), then show that
( )
( )
Vn e jω = e−jωkXn e jω .
6
4. a) Define complex cepstrum and real cepstrum of a sequence x (n). Show that the complex cepstrum of a sequence can be computed as,
xˆ (n) =
−1 2πnj
1 xˆ (0) = 2π
∫
∫
π
−π
π
x′(e jω ) jωn e dω, n ≠ 0 & x(e j ω )
log X(e jω ) dω
−π
where symbols have usual meanings.
6
*PED136*
PED – 136
-3-
b) Find the cepstrum of the sequence whose Z-transform is V(z) =
1+ 0.98z−1
π π j −j −1 6 6 (1 − 0.9e z )(1 − 0.9e z−1)
.
6
c) Explain with a block diagram steps involved in computing the real cepstrum of a speech signal. Draw block diagram to explain how pitch of the speech signal is estimated from the computed cepstrum and justify the steps.
8
5. a) Define a pth order linear predictor. Show that the following relation holds for a pth order linea predictor s (n)
p
∑ k =1αks(n − k) + e(n),
where { α k} are LP
coefficients and e (n) is the prediction error.
4
b) Explain the autocorrelation method of LP analysis and show that i)
p
∑ k =1α k Rn ( i− k ) = Rn (i),
1 ≤ i ≤ p, and
assuming p = 4, write the above equation in the matrix form. ii) minimum mean squared prediction error En = Rn(0) − ∑ pk =1 αkRn (k) .
8
c) Using Durbin’s recursive algorithm compute the transfer function of 2nd order LP model for a speech signal having autocorrelation coefficients R (0) = 1, R (1) = 0.5 and R (2) = 0.25.
8
6. a) Explain the following : i) JND, ii) frequency masking and iii) temporal masking. b) Discuss two broad classes of speech quality measures.
12 8
PED – 136
-4-
*PED136*
7. a) With a block diagram explain ADPCM system with feed-forward adaptive quantization.
6
b) Discuss the salient features of the LD_CELP G.728 speech compression standard.
6
c) With a block diagram explain MPEG audio compression algorithm.
8
8. a) Explain the following in the context of speech signal processing. i) Dynamic time warping and ii) Hidden Markov model. b) Write a short note on Text to speech synthesis.
... Enâ1 + x2 (n),where the window h (n), is defined as h (n) = anu(n), x (n) is. the speech signal, and 0 < a < 1. Draw a digital network diagram of this equation. 6.
5. What are diphthongs? 6. Why we require sampling of analog signals? 7. Why we generally keep sampling frequency more than Nyquist rate for sinusoidal. signals? 8. What are the assumptions uses while designing a uniform quantizer? 9. What are the ad
as VisualStudio, GCC, Eclipse, Pycharm and has undertaken audio signal processing course. Page 1 of 1. JD_ Audio signal processing.pdf. JD_ Audio signal ...
from any use or operation of any methods, products, instructions, or ideas contained in the material herein. British Library Cataloguing in Publication Data.
Sep 14, 2007 - recording is too high for many computer applications. In addition, the digital bit rates .... based entertainment, training, and demonstration systems. Over the course of the next ... over a computer network. Until recently, the use ..
Sep 14, 2007 - inputted to the analysis filter bank. The di?'erences between the reconstructed audio sound track and the original sound track can be made ...
on each new speaker approaching that of an SD system for that speaker, while ... over the telephone, one can only count on a few seconds of unsupervised speech. ... The authors are with the Panasonic Speech Technology Laboratory, Pana-.
voice approach with other speaker adaptation algorithms, the ...... the 1999 International Conference on Acoustics, Speech, and Signal Processing. He was an ...
We are looking for two doctoral students for the ICHO project (Immersive Concerts for. Homes). The aim of the project is to bring the immersive concert experience to people's homes with the help of head-tracked headphones and sophisticated signal pro
performance for clean and noisy images but also audio-visual speech recognition ..... [4] Ross, L. A., Saint-Amour, D., Leavitt, V. M., Foxe, J. J. Do you see what I ...
Email: {asubram, sabrig, epatter, jgowdy}@clemson.edu. Abstract. In recent years ... visual information for the purposes of Automatic Speech. Recognition (ASR) ...
Department of Electrical and Computer Engineering. Clemson University ..... Interaction,â PhD Dissertation, University of Illinois, Urbana-. Champaign, 1999.