Vector quantization (VQ)

Vector quantization (VQ) is a classical quantization technique from signal processing which allows the modeling of probability density functions by the distribution of prototype vectors. It was originally used for data compression. It works by dividing a large set of points (vectors) into groups having approximately the same number of points closest to them. Each group is represented by its centroid point, as in k-means and some other clustering algorithms.

Vector quantization

From Wikipedia, the free encyclopedia
  (Redirected from Vector Quantization)

Vector quantization (VQ) is a classical quantization technique from signal processing which allows the modeling of probability density functions by the distribution of prototype vectors. It was originally used for data compression. It works by dividing a large set of points (vectors) into groups having approximately the same number of points closest to them. Each group is represented by its centroid point, as in k-means and some other clustering algorithms.

The density matching property of vector quantization is powerful, especially for identifying the density of large and high-dimensioned data. Since data points are represented by the index of their closest centroid, commonly occurring data have low error, and rare data high error. This is why VQ is suitable for lossy data compression. It can also be used for lossy data correction and density estimation.

Vector quantization is based on the competitive learning paradigm, so it is closely related to the self-organizing mapmodel.

 

Training[edit]

A simple training algorithm for vector quantization is:

  1. Pick a sample point at random
  2. Move the nearest quantization vector centroid towards this sample point, by a small fraction of the distance
  3. Repeat

A more sophisticated algorithm reduces the bias in the density matching estimation, and ensures that all points are used, by including an extra sensitivity parameter:

  1. Increase each centroid's sensitivity by a small amount
  2. Pick a sample point at random
  3. Find the quantization vector centroid with the smallest <distance-sensitivity>
    1. Move the chosen centroid toward the sample point by a small fraction of the distance
    2. Set the chosen centroid's sensitivity to zero
  4. Repeat

It is desirable to use a cooling schedule to produce convergence: see Simulated annealing.

The algorithm can be iteratively updated with 'live' data, rather than by picking random points from a data set, but this will introduce some bias if the data is temporally correlated over many samples. A vector is represented either geometrically by an arrow whose length corresponds to its magnitude and points in an appropriate direction, or by two or three numbers representing the magnitude of its components.

Applications[edit]

Vector quantization is used for lossy data compression, lossy data correction, pattern recognition and density estimation.

Lossy data correction, or prediction, is used to recover data missing from some dimensions. It is done by finding the nearest group with the data dimensions available, then predicting the result based on the values for the missing dimensions, assuming that they will have the same value as the group's centroid.

For density estimation, the area/volume that is closer to a particular centroid than to any other is inversely proportional to the density (due to the density matching property of the algorithm).

Use in data compression[edit]

Vector quantization, also called "block quantization" or "pattern matching quantization" is often used in lossy data compression. It works by encoding values from a multidimensional vector space into a finite set of values from a discretesubspace of lower dimension. A lower-space vector requires less storage space, so the data is compressed. Due to the density matching property of vector quantization, the compressed data has errors that are inversely proportional to density.

The transformation is usually done by projection or by using a codebook. In some cases, a codebook can be also used toentropy code the discrete value in the same step, by generating a prefix coded variable-length encoded value as its output.

The set of discrete amplitude levels is quantized jointly rather than each sample being quantized separately. Consider a k-dimensional vector [x_1,x_2,...,x_k] of amplitude levels. It is compressed by choosing the nearest matching vector from a set of n-dimensional vectors [y_1,y_2,...,y_n], with n < k.

All possible combinations of the n-dimensional vector [y_1,y_2,...,y_n] form the vector space to which all the quantized vectors belong.

Block Diagram: A simple vector quantizer is shown below

Only the index of the codeword in the codebook is sent instead of the quantized values. This conserves space and achieves more compression.

Twin vector quantization (VQF) is part of the MPEG-4 standard dealing with time domain weighted interleaved vector quantization.

Video codecs based on vector quantization[edit]

The usage of video codecs based on vector quantization has declined significantly in favor of those based on motion compensated prediction combined with transform coding, e.g. those defined in MPEG standards, as the low decoding complexity of vector quantization has become less relevant.

Audio codecs based on vector quantization[edit]

Use in Pattern Recognition[edit]

VQ was also used in the eighties for speech[3] and speaker recognition.[4] Recently it has also been used for efficient nearest neighbor search [5] and on-line signature recognition.[6] In pattern recognition applications, one codebook is constructed for each class (each class being a user in biometric applications) using acoustic vectors of this user. In the testing phase the quantization distortion of a testing signal is worked out with the whole set of codebooks obtained in the training phase. The codebook that provides the smallest vector quantization distortion indicates the identified user.

The main advantage of VQ in pattern recognition is its low computational burden when compared with other techniques such as dynamic time warping (DTW) and hidden Markov model (HMM). The main drawback when compared to DTW and HMM is that it does not take into account the temporal evolution of the signals (speech, signature, etc.) because all the vectors are mixed up. In order to overcome this problem a multi-section codebook approach has been proposed.[7] The multi-section approach consists of modelling the signal with several sections (for instance, one codebook for the initial part, another one for the center and a last codebook for the ending part).

See also[edit]

Part of this article was originally based on material from the Free On-line Dictionary of Computing and is used withpermission under the GFDL.

References[edit]

  1. Jump up^ "Bink video"Book of Wisdom. 2009-12-27. Retrieved 2013-03-16.
  2. Jump up^ "Vorbis I Specification". Xiph.org. 2007-03-09. Retrieved 2007-03-09.
  3. Jump up^ Burton, D. K.; Shore, J. E., Buck, J. T. (1983). "A generalization of isolated word recognition using vector quantization".IEEE International Conference on Acoustics Speech and Signal Processing ICASSP: 1021–1024.
  4. Jump up^ Soong, F.; A. Rosenberg, L. Rabiner, B. Juang (1985). "A vector Quantization approach to Speaker Recognition". IEEE Proceedings International Conference on Acoustics, Speech and Signal Processing ICASSP 1: 387–390.
  5. Jump up^ H. Jegou; M. Douze and C. Schmid (2011). "Product Quantization for Nearest Neighbor Search". Transactions on Pattern Analysis and Machine Intelligence 33 (1): 117–128. doi:10.1109/TPAMI.2010.57.
  6. Jump up^ Faundez-Zanuy, Marcos (2007). "On-line signature recognition based on VQ-DTW"Pattern Recognition 40 (3): 981–992.
  7. Jump up^ Faundez-Zanuy, Marcos; Juan Manuel Pascual-Gaspar (2011). "Efficient On-line signature recognition based on Multi-section VQ"Pattern Analysis and Applications 14 (1): 37–45.

External links[edit]

RELATED ARTICLESExplain
Machine Learning Methods & Algorithms
Unsupervised learning
Vector quantization (VQ)
Linde–Buzo–Gray algorithm
Association rule learning
Data clustering
Expectation–maximization algorithm
FastICA
Generative topographic map
Hierarchical clustering
IBSEAD - distributed autonomous entity systems based Interaction
Information bottleneck method
Partitional clustering
Radial basis function network
Self-organizing map
Sparse PCA (sparse principal component analysis)
Stochastic gradient descent
Graph of this discussion
Enter the title of your article


Enter a short (max 500 characters) summation of your article
Enter the main body of your article
Lock
+Comments (0)
+Citations (0)
+About
Enter comment

Select article text to quote
welcome text

First name   Last name 

Email

Skip