Question: Explain the Levinson Durbin Algorithm for calculation of prediction coefficients
0

Subject: Speech Processing

Topic: LPC and Parametric Speech Coding

Difficulty: Medium

sp(38) • 440 views
 modified 8 months ago by written 10 months ago by awari.swati831 • 150
0

L-D Algorithm:

(i) Take 128 samples of speech signals.

(ii) Find the value for auto-correlation coefficients for segment length equal to 128 - |l - k|, where (l-k) denotes the offset. We use the formula to find the normalized auto-correlation given by: $$\gamma_{xx}(i) = \frac{\sum_{i=1}^{N-i} x(k)x(k+i)}{\sqrt{\sum_{i=1}^{N-i} x(k) x(k) + \sum_{i=1}^{N-i} x(k-i) x(k-i)}}$$

(iii) To find the reflection coefficient, find error after each iteration.

(iv) Use recursive to find filter coefficient for further higher value of prediction error.