.\" Copyright (c) 1987 Entropic Speech, Inc.; All rights reserved.\" @(#)classify.3	1.3 30 Apr 1997 ESI.TH CLASSIFY 3\-ESPSsp 30 Apr 1997.ds ]W "\fI\s+4\ze\h'0.05'e\s-4\v'-0.4m'\fP\(*p\v'0.4m'\ Entropic Speech, Inc..SH NAMEclassify \- maximum-likelihood or maximum-posterior-probability classificationof feature vector.SH SYNOPSIS.ft B.nfintclassify(feavec, nfea, nclas, means, invcovars, priors, posteriors)float	*feavec;int	nfea, nclas;float	**means, ***invcovars, *priors;float	*posteriors;.fi.ft.SH DESCRIPTION.PP.I Classifyclassifies a vector.I feavecof.I nfeanumerical features into one of.I nclasclasses, where each class is represented by an.IR nfea -elementmean vector, an.IR nfea -by- nfeainverse covariance matrix, and, for maximum-posterior-probabilityclassification, a prior probability..PPThe input arrays.I meansand.I invcovarsshould each have.I nclaselements, as should.I priorsfor maximum-posterior-probability classification; formaximum-likelihood classification,.I priorsshould be NULL.For each index.I cin the range.RI "(0 <= " c " < " nfea "),".IR means [ c ]and.IR invcovars [ c ]should point to the mean vector and inverse covariance matrix for oneclass, and.IR priors [ c ],if defined, should be the prior probability for the same class.(More precisely,.IR means [ c ]points to the first element of the mean vector, and.IR invcovars [ c ]points to a pointer to the first row of the inverse covariance matrix.).PP.I Posteriorsshould point to the first element of an.IR nfea -elementoutput vector, which will receive the posterior probabilities.If.I priorsis NULL, these are computed as though all priors had been specified as.RI 1/ nclas..PPThe return value of.I classifyis the index.I cfor which the posterior probability is greatest.How ties are broken is unspecified..SH BUGSNone known..SH "SEE ALSO"\fIfea_stats\fR(1\-ESPS),  \fIf_mat_alloc\fR(3\-ESPSu), FEA_STAT(5\-ESPS).SH AUTHORRodney Johnson