Help in LBG Vector Quantization - Splitting Factor

Signal Processing Asked by an6 on January 13, 2021

I’m currently making a program for speech recognition. In the step of codebook generation using the LBG (Linde-Buzo-Gray) algorithm, I’ve read that the splitting factor $$varepsilon = 0.01$$ (generally)

The splitting factor is used to split the centroid of the speech features according to the formulae

begin{align} Y_{n}^+ &= Y_n (1+varepsilon)\ Y_{n}^- &= Y_n (1-varepsilon) end{align}

where $$n$$ is the index of the given codeword/centroid to be split and $$Y_n$$ is the codeword.

Also, after the codebook is generated, nearest neighbours are searched for each speech feature vector and the centroids are updated (basically clustering of features). This is done until the distortion of the codebook is less than epsilon.

Although my program seems to be working fine, I’m interested to know why the splitting factor is usually set to be 0.01.

Any help is appreciated. This is my first time working with codebooks and vector quantization.

Related Questions

Detecting corners using structure tensor matrix

1  Asked on November 25, 2021 by nagabhushan-s-n

How to Extrapolate a 1D Signal?

4  Asked on November 23, 2021

Remove Shadows from Contour of an Object (Grapes)

1  Asked on November 23, 2021 by alon-shmiel

How to Extrapolate a 1D Chirped Signal?

0  Asked on November 23, 2021

Problem converting from Eb/N0 to SNR

1  Asked on November 23, 2021 by researcher9

How to calculate variance of noise on a noisy image (Noise+Signal)

1  Asked on November 23, 2021 by image-check

Group delay plot of a Moving average filter

1  Asked on November 23, 2021

Hello World Example of a Digital Waveguide in JavaScript

0  Asked on November 23, 2021

Confusion calculating length of Kaiser window

1  Asked on November 21, 2021

Algorithms to re-spatialize a stereo recording audio signal?

5  Asked on November 21, 2021 by g6kxjv1ozn

What could cause fast Fourier transform to give complex conjugate of the intended result?

2  Asked on November 21, 2021

Derivation of the LMMSE (Linear Minimum Mean Squared Error) Estimate and the MMSE Under Gaussian Prior

2  Asked on November 21, 2021 by mczhang

signed 2’s complement of a negative number for Integer = 3 and Fraction = 9

1  Asked on November 21, 2021

Efficient double upsampling of a pure real tone

2  Asked on November 15, 2021 by cedron-dawg

Frequency range for STFT in Librosa

1  Asked on November 13, 2021 by cralle

Sampling period

1  Asked on November 13, 2021 by john_hb

Trying to find an intuitive understanding of OQAM/FBMC

1  Asked on November 11, 2021 by ali-khalil

Eroding noisy borders of polygons in an image

1  Asked on November 11, 2021

U-nets : how exaclty is upsampling performed

1  Asked on November 8, 2021

Separating/recovering base signal from two mixed signals, given phase information

1  Asked on November 8, 2021 by rotano