Subject: Speech Processing

Topic: Homomorphic Speech Processing

Difficulty: Low

(i) Cepstral domain approach for pitch period measurement.

(ii) A cepstral is obtained when we take the Fourier Transform (FT) of log spectrum.

(iii) The speech signal frame consisting of 256 samples as inputs.

(iv) The signal is passed through a Hamming window which is preferred over rectangular counterpart because of its smoother edge which suppresses the Gibbs phenomenon.

(v) Window signals passed through FFT section, then log amplitude signals is obtained.

(vi) IFFT transform to compute the cepstral coefficient of the speech frame.

(vii) The type of system which converts convolved components into additive one is called ahomomorphic system for convolution.