What is PLP? Compare the procedure to calculate MFCC to PLP

28views

written 6.0 years ago by

Block schematic for perceptual linear prediction

(i) Estimation of power spectrum: For analysis window. Here, the signal is passed through a Hamming window and the squared magnitude of FFT is taken as window.

(ii) Critical Band Analysis power spectrum is integrated within the critical band filter response.

(iii) Equal loudness pre-emphasis compensate for the non-equal perception of loudness at different frequencies.

(iv) Intensity loudness conversion log taken before integration. This compress the power spectrum.

(v) Auto-regressive model is used to smooth the spectrum.

Sr.	MFCC	PLP
1.	Calculates power spectrum	Calculates power spectrum
2.	Integration of power spectrum using Mel filter bank	Integration of power spectrum using Bark scale filter bank
3.	Pre-emphasis done before spectral analysis	Pre-emphasis is done using equal loudness pre-emphasis curve
4.	Log is taken before integration for spectral compression	Cube root of power spectrum is taken
5.	Take IDFT, get cepstral coefficients	Take IDFT, get auto-correlation
6.	Spectral smoothing using cepstral transaction	Auto-regressive model is used to smooth the spectrum
7.	No such option is present	Coefficients converted to cepstral coefficient optionally
8.	Lifting is done to reduce variances of cepstral coefficients	Lifting is done to reduce variances of cepstral coefficients

ADD COMMENT EDIT