SPEECH AND AUDIO SIGNAL PROCESSING.pdf - Google Drive

’

*JEP1056*

JEP – 1056

II Semester M.E. (Electronics & Communication) Degree Examination, July 2014 (2K9 Scheme) EL 215.4 : Elective – I : SPEECH AND AUDIO SIGNAL PROCESSING Time : 3 Hours

Max. Marks : 100

Instruction : Answer any five full questions. 1. a) Write a block diagram to illustrate speech production mechanism. Discuss the significance of source and filter in the context of producing vowels, fricatives and plosives sounds. b) Write a note on : Spectrograms.

8 6

c) Discuss the following speech sounds in terms of source, filter, places and manner of articulation : i) Nasals and ii) Diphthongs. Give an example for each.

6

2. a) Define short time energy and short time average zero crossing rate.

4

b) Discuss speech Vs silence discrimination algorithm based on short time energy and short time average zero crossing rate. Indicate the limitation of the technique.

8

c) Discuss the significance of short time autocorrelation function and 3 level central clipper in the context of speech signal processing.

8

3. a) Explain short time Fourier transform and give its filter bank interpretation in terms of low pass and band pass filters.

6

b) Discuss the overlap and add method of short time synthesis and derive the expression for necessary constraints.

8

c) Show that short time autocorrelation function and short time psd form a Fourier transform pair.

6 P.T.O.

JEP – 1056

-2-

*JEP1056*

4. a) State and prove any two properties of complex cepstrum by considering rational z-transform of the form. |A|∏

X (z) =

(∏

Ni

Mi k =1

k =1

6

(1 − a kz −1) ∏M0 (1 − bkz) k =1

. (1 − ckz −1)) (∏N0 (1 − dkz)) k =1

b) Find the complex and real cepstrum of the sequence p(n) = δ(n) +

1 1 δ(n − 25 ) + δ(n − 50 ) . 2 4

c) Explain with a block diagram steps involved in computing the real cepstrum of a speech signal. Draw a block diagram to explain how formants of the speech signal are estimated from the computed cepstrum and justify the steps.

6

8

5. a) Define a pth order linear predictor. Show that the following relation holds for a pth order linear predictor s (n) =

p

∑k=1aks(n − k) + e(n), where {ak} are LP

coefficients and e (n) is the prediction error.

4

b) Explain the autocovariance method of LP analysis and show that i)

p

∑k =1α kφn( i − k ) = φn (i, 0),

1 ≤ i ≤ p, and assuming p = 4, write the above

equation in the matrix form. ii) the diagonal elements of the matrix are given by, φn(i + 1, k + 1) = φn(i, k) + sn(− i − 1)sn(− k − 1) − sn(N − 1 − i)sn (N − 1 − k ) .

8

c) Using Durbin’s recursive algorithm compute the transfer function of 2nd order LP model for a speech signal whose autocorrelation sequence is R(k)= (24/5) × 2–|k| – (27/10) × 3–|k|. 6. a) Explain the concept of masking in the context of speech perception.

8 6

b) Explain how VQ techniques can be applied to classify speech vectors.

8

c) Explain 3 different groups of speech coders.

6

*JEP1056*

-3-

JEP – 1056

7. a) With a block diagram explain ADPCM system with feed-forward adaptive quantization. b) Write a note on : i) LD_CELP G.728 speech compression standard. ii) MPEG audio compression algorithm.

6

(6+8)

8. Write short notes on : i) Mel frequency cepstral coefficients. ii) Hidden Markov model, and iii) Text to speech synthesis. __________________

(6+6+8)