Browse Prior Art Database

Adaptive Filtering of Speech Signals

IP.com Disclosure Number: IPCOM000062374D
Original Publication Date: 1986-Nov-01
Included in the Prior Art Database: 2005-Mar-09
Document File: 2 page(s) / 51K

Publishing Venue

IBM

Related People

Gerhardt, LA: AUTHOR [+2]

Abstract

A method is described for automatically controlling an adaptive filter bank used for speech processing. The use of adaptive filters allows the spectral characteristics of the incoming speech signal to be estimated using fewer filters than would be required with a stationary filter bank system. Referring to Fig. 1, in the prior art, the incoming speech signal 10 would be processed by a filter bank consisting of a number of stationary bandpass filters 20. The output of each of the filters 20 is processed by an energy detector 30, providing a fixed bank spectral estimate. The disadvantage of this approach is that many filters are required to capture the spectral characteristics accurately enough for applications such as speech recognition.

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 57% of the total text.

Page 1 of 2

Adaptive Filtering of Speech Signals

A method is described for automatically controlling an adaptive filter bank used for speech processing. The use of adaptive filters allows the spectral characteristics of the incoming speech signal to be estimated using fewer filters than would be required with a stationary filter bank system. Referring to Fig. 1, in the prior art, the incoming speech signal 10 would be processed by a filter bank consisting of a number of stationary bandpass filters 20. The output of each of the filters 20 is processed by an energy detector 30, providing a fixed bank spectral estimate. The disadvantage of this approach is that many filters are required to capture the spectral characteristics accurately enough for applications such as speech recognition. If adaptive filters are used, the filters can move into the high energy regions of the spectrum and capture these regions accurately. The high energy regions carry the bulk of the information, and the position of the spectral peaks varies with time. The adaptive system is able to track this movement and extract the important spectral information efficiently. Fig. 1 shows the addition of an adaptation control block 40 which takes the energy readings from each of the filters, constructs a cumulative energy vs. frequency estimate and repositions the filters. The objective of the repositioning is to divide the spectral energy equally among the filters, thereby concentrating the filters in the high...