At present the most popular criteria are of informative heuristic criteria associated with the estimation of separability given classes and based on the fundamental pattern recognition compactness hypothesis: with increasing distance between the classes improved their separability. “Good” are those features that maximize the relationship. Such heuristic criteria, although are widely used in solving practical problems of classification, but in theory are scarcely explored. At present, the method of selecting informative features, taking into account the relationships of features based on heuristic criteria, has not been developed. The report considers this task.
The definition of an informative set of features is one of the important tasks in pattern recognition. Typically, the determination of informative features is carried out using two types of methods. The first type of methods is “direct methods”, they are directly aimed at identifying informative sets of attributes. And the second type is called "inverse methods", these methods serve to build informative sets of signs by eliminating uninformative signs from the attribute space. This article is devoted specifically to the development of the second type of method; it proposes an accelerated method and an algorithm for determining non-informative features based on the selected non-informative criterion.
Abstract: In this paper, the systems of speaker identification of a text-dependent and independent nature were considered.
Feature extraction was performed using chalk-frequency cepstral coefficients (MFCC). The vector quantization method for the automatic identification of a person by voice has been investigated. Using the extracted features, the code book from each speaker was built by clustering the feature vectors. Speakers were modeled using vector quantization (VQ). Using the extracted features, the code book from each speaker was built by clustering the feature vectors. Codebooks of all announcers were collected in the database. From the results, it can be said that vector quantization using cepstral features produces good results for creating a voice recognition system.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.