54213-mt----speech Processing

Code No: 54213/MT

NR M.Tech. – II Semester Regular Examinations, September, 2008 SPEECH PROCESSING (Common to Digital Systems & Computer Electronics/ Digital Electronics & Communication Systems) Time: 3hours Max. Marks:60 Answer any FIVE questions All questions carry equal marks --1.a) How the architecture (speech organs) are positioned for producing (i) Nasals (ii) Stop consonants (iii) Fricatives b) Give a brief note about digital modelling of speech signal. 2.a) b) 3.a) b)

What is the significance of zero crossing rate, short time energy and average magnitude. How do these values are determined. List and explain the methods for pitch period estimation using time domain methods. Explain the steps involved in the autocorrelation method to compute LPC coefficient. Explain how the vocal tract system is estimated using LPC method.

4.a) b)

Explain the properties of complex cepstrum. Explain how the formants are estimated from the complex cepstrum of speech.

5.a)

Give a brief note about various categories of speech enchancement techniques. Explain the re – synthesis method followed for speech enhancement.

b) 6.a) b) 7.a) b) 8.a) b)

What is meant by parametric representation of speech? What are its advantages over non – parametric representations. Explain the step – wise procedure followed for implementing isolated digit recognition system with a typical flow diagram. Explain the steps involved in Baum – Welch re – estimation method of HMM training. What is language model? How is it benificial is improving speech recognition performance. Give a brief note about the feactures that distinguish speakers. Explain with a suitable flow diagram, a typical speaker recognition system. &_&_&_&