Robust codebooks for vector quantization
AssigneeSharp Laboratories of America, Inc.
InventorsRenjit Tom Thomas, Shawmin Lei
Life of this patent
Estimating the term…
Abstract
A method for data compression. An encoder receives data vectors from an original data set. The encoder uses a vector quantization codebook to encode the data vectors into encoded vectors. The codebook is constructed from a compound data set, where the compound data set includes real data vectors and artificial data vectors. The encoded vectors are indexed in the codebook and the indexes are transmitted across communication channels or transmitted to storage.
Claims17
A method of data compression, the method comprising:
- a) receiving input vectors representative of an original data set to be compressed;
- b) encoding the input vectors using a vector quantization codebook to produce encoded vectors, wherein the codebook is produced from a compound data set including:
- i) real data;
- and
- ii) at least one set of artificial data;
- c) transmitting the indexes of the encoded vectors.
An encoder, the encoder configured to:
- a) receive input vectors from the original data set to be compressed;
- b) encode the input vectors using a vector quantization codebook to produce encoded vectors, wherein the codebook is produced from a compound data set including:
- i) real data;
- and
- ii) at least one set of artificial data;
- c) transmit the indexes of the encoded vectors.
A method of producing a VQ codebook, the method comprising:
- a) defining an initial codebook having code vectors;
- b) assigning training vectors from a training set to code vectors in the training set, wherein the training set includes both real data and artificial data;
- c) calculating a centroid for a set formed from the code vectors assigned to a particular code vector;
- d) replacing the particular code vector with the centroid;
- e) determining if error has been reduced more than a predetermined threshold;
- and
- f) repeating assigning, calculating and replacing if the error has been reduced more than the predetermined threshold.
An article containing machine-readable code that, when executed, causes the machine to:
- a) define an initial codebook having code vectors;
- b) assign training vectors from a training set to code vectors in the training set, wherein the training set includes both real data and artificial data;
- c) calculate a centroid for a set formed from the code vectors assigned to a particular code vector;
- d) replace the particular code vector with the centroid;
- e) determine if error has been reduced more than a predetermined threshold;
- and
- f) repeat assigning, calculating and replacing if the error has been reduced more than the predetermined threshold.
Dependencies are inferred from the claim text.