Scribd
Upload
42 views

Speech and Audio Processing: Lecture-4

The document discusses methods for determining pitch period in speech signals. It describes how earlier methods can only find integer pitch periods while higher resolution is often needed. It introduces the Medan, Yair, and Chazan algorithm for finding fractional pitch periods. It also discusses finding the optimal integer-valued pitch period by minimizing the normalized sum of squared error between two consecutive speech frames. The optimal fractional pitch period is defined as the continuous-valued pitch period that is the optimal integer period plus a fractional component.

Uploaded by

Randeep Singh
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
42 views

Speech and Audio Processing: Lecture-4

The document discusses methods for determining pitch period in speech signals. It describes how earlier methods can only find integer pitch periods while higher resolution is often needed. It introduces the Medan, Yair, and Chazan algorithm for finding fractional pitch periods. It also discusses finding the optimal integer-valued pitch period by minimizing the normalized sum of squared error between two consecutive speech frames. The optimal fractional pitch period is defined as the continuous-valued pitch period that is the optimal integer period plus a fractional component.

Uploaded by

Randeep Singh
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 20

Speech and Audio Processing

Lecture-4 By: Mohit Goel


[email protected]

Fractional Pitch Period


The methods discussed earlier can only find integervalued pitch periods. That is, the resultant period values are multiples of the sampling period 1/(8 kHz)=0.125 ms. In many applications, higher resolution is necessary to achieve good performance. For that Medan, Yair, and Chazan Algorithm is used.

Optimal Integer-Valued Pitch Period Consider a speech frame that ends at time instant n = m, with a length of N. The frame can be expressed by

e[n] is error signal. Note from Figure that two consecutive frames of length N are involved. The optimal pitch period at time instant n = m can be defined as the particular value of N, denoted by N0, that minimizes the normalized sum of squared error

The optimal value of b can be found by differentiating J with respect to b and setting the result to zero. This gives

Put value of b in above equation we will get result:

Optimal Fractional Pitch Period Consider the continuous-valued pitch period T0 defined by where Ts is the sampling period 1/(8 kHz) = 0.125 ms and is the fractional pitch period.

Also

You might also like