Browse Prior Art Database

Determination of Expected Probability of Histograms of a Given Length in the Polling Fast Match

IP.com Disclosure Number: IPCOM000036896D
Original Publication Date: 1989-Nov-01
Included in the Prior Art Database: 2005-Jan-29
Document File: 2 page(s) / 24K

Publishing Venue

IBM

Related People

Bahl, LR: AUTHOR [+4]

Abstract

A technique is described whereby algorithms are used in speech recognition devices to determine the expected probability of histograms of a given length in the polling fast match. It allows the comparison of probabilities of histograms of different lengths, in a meaningful way, during a polling fast match operation [*].

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 54% of the total text.

Page 1 of 2

Determination of Expected Probability of Histograms of a Given Length in the Polling Fast Match

A technique is described whereby algorithms are used in speech recognition devices to determine the expected probability of histograms of a given length in the polling fast match. It allows the comparison of probabilities of histograms of different lengths, in a meaningful way, during a polling fast match operation [*].

Given a word, the Poisson polling fast match provides a method for computing the probability of any given histogram, H=(h1, h2, ..., hn), of the n fenemes, f1, f2, ...,fn, that are possible as output from an acoustic processor. Specifically, where gi is the expected number of occurences of fi in an utterance

The probability of a histogram decreases rapidly as its length increases. In order to make a meaningful comparison of the probabilities of historgrams of different lengths, it is important to be able to compare the probability of a histogram to the expected probability of histograms of the same length. The concept described herein allows the expected probability of a histogram to be computed as a function of its length. Let gj(l,k) = Pr(hl=0,i<j; hj=k; L(H)=l).

In other words, gj(l,k) is the probability that hk is the first non-zero h, that hj=K, and that the histogram has length l. The probability can be expressed such that a histogram will have length l in terms of g1 . As a result, But, the gj obey the following relations and finally,

The gj(l,k) have a large dynamic range, and so it is importa...