Browse Prior Art Database

A Method For Calibrating Speech Recognition Testing In Small Audio Chambers Disclosure Number: IPCOM000030306D
Original Publication Date: 2004-Aug-04
Included in the Prior Art Database: 2004-Aug-04
Document File: 2 page(s) / 89K

Publishing Venue



When testing speech engines in a small audio test enclosure with a speaker and a microphone, the sound reflections inside the enclosure can affect the test results negatively. By setting an optimal microphone gain, speaker loudness and the distance between them, this negative effect can be minimized to yield more consistent and reliable test results

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 52% of the total text.

Page 1 of 2

A Method For Calibrating Speech Recognition Testing In Small Audio Chambers

Disclosed is a method to develop speech recognition test setups to yield more consistent and reliable test results. A general speech recognition test setup consists of: a PC, a speaker, a microphone and a small audio test enclosure. Several of these test setups can be placed near each other in a test area. Within each test setup, the PC is used to run a Speech Engine and Test Control Program; The speaker is controlled and used to simulate a person speaking by the Test Control Program; The microphone is controlled and used as an input device by the Speech Engine; The small audio test enclosure is used as a sound insulation chamber. Following is an example of a test setup.

The objective is to minimize the speaker output reflections within the test enclosure. It was proven that the sound reflections is the main cause for poor test results provided that the same speaker, microphone and software were used. Several setup properties can be used to minimize this problem. Since every microphone, speaker and audio test enclosure behaves differently depends on microphone, speaker models and enclosure shape/size, An optimized setup is different for each application or component inside the test enclosure. The optimization can only be achieved through trial and error but within certain guidelines. Following are the guidelines that were used to optimize a test setup.

1) The lower the speaker level, the lower the reflections. By applying this guideline, the speaker should be set as low as possible. However, if the speaker is set too low, the sound quality can be deteriorated. This level should be experimented along with other


[This page contains 1 picture or other non-text object]

Page 2 of 2

settings to obtain an optimized level.

2) The closer the microphone, the stronger the received audio signal. By applying this guideline, the microphone should be placed as close to the speaker as possible. However, the speaker is gener...