Using Voice to Solve Ergonomic Problems Dr. William Lenharth, CHFP UNH – Project54.

Post on 19-Jan-2016

217 views 0 download

Transcript of Using Voice to Solve Ergonomic Problems Dr. William Lenharth, CHFP UNH – Project54.

Using Voice to Solve Ergonomic Problems

Dr. William Lenharth, CHFP

UNH – Project54

When to use Voice?• A high level of visual activity is present.

• When hands are busy with critical tasks.

• If speeds match …commands - activity

Problems with Voice• Training

• Accuracy

• Noise environment

• Required hardware

Solutions - Hardware• Hardware: Microphone types

- Environment driven:

- Noise Cancelling

- Headset

- Throat

Project54 Mobile Platform

Hardware ConnectivityHardware Connectivity

Solution - Accuracy• How to improve accuracy?

- Voice engines use a statistical approach.

- Limit correct answers by changing grammar

e.g.: If the question is, Sex? Limit grammar

to “male” or “female”.• Grammar is a text file easy to change.

Sample Grammar

•  [<radar>]• <radar> = front antenna "FRONT ANTENNA"• <radar> = front antenna off "FRONT ANTENNA OFF"• <radar> = rear antenna "REAR ANTENNA"• <radar> = rear antenna off "REAR ANTENNA OFF"• <radar> = radar "RADAR"• <radar> = lock "LOCK"• <radar> = lock off "LOCK OFF"

SOFTWARE ARCHITECTURE

Speechrecognition

engine

Grammar

recognitionresults

Commandlist

Application 1

Application 2Application n

Text-to-speechengine

coordinatedspeech output

speechoutput

Initializationlist

audioobject

feedback

Application manager

Speechoutput

coordinator

Loggingcoordinator

Messagecoordinator

Keyboard andmouse input

loggingmessages

controlmessagesstatus

messages

Solution - Training • Voice systems: Speaker Dependent or

independent.

• Voice based allows 20% variation max.

• Systems can be trained with a file to help improve accuracy.

System - Control• Push to Talk : Controls when the voice is

being processed by the computer.

• Continuous speech such as dictation.

Project54• Nine years experience

• 1200+ vehicles in use daily Nationwide

• Inexperienced users - over 60,000 voice samples collected.

Project54 – Voice System• Speaker Independent

• Push-to-talk controlled

• Grammars changed “on the fly”

• 96% accuracy using free speech engine

• Noise cancelling microphone

How build your Application

• Use Project54 base software to handle speech and process control.

• Build your own application using P54 GUI and SUI support.

• Connect your control computer to your device via the P54 CAN2B bus and interconnect devices (or other).

• Deploy system and test …

Further Information

• WWW.project54.unh.edu

• whl@sr.unh.edu

15