Multiple Input/Output Connector - PowerPoint PPT Presentation

About This Presentation
Title:

Multiple Input/Output Connector

Description:

Speech can be converted to text. Palm Pilot graffiti can be ... Implicit input such as body gestures and facial expressions cannot be converted to text ... – PowerPoint PPT presentation

Number of Views:59
Avg rating:3.0/5.0
Slides: 17
Provided by: saga7
Learn more at: http://www.ai.mit.edu
Category:

less

Transcript and Presenter's Notes

Title: Multiple Input/Output Connector


1
Multiple Input/Output Connector
  • Shalini Agarwal
  • Oxygen Conference
  • January 8th, 2002

2
Multiple Input/Output Connector
  • Shalini Agarwal
  • Oxygen Workshop, January, 2002

3
Agenda
  • Goals
  • Desired Functionality
  • Valid Inputs/Outputs
  • System Overview
  • Multiple Output Demo
  • Future Work

4
Goals
  • Scenario Multiple input devices multiple
    applications
  • Challenge Route input to correct application
  • Flexible design
  • Dynamically add input devices (and output)
  • Basic socket design to allow for different
    applications
  • Transparent use of different inputs and outputs

5
We want three types of functionality
  • Single input, Multiple output
  • Multiple input, Single output
  • Multiple input, Multiple output

6
What kind of input can we accommodate?
  • Not all input modes are compatible, but some are
  • Speech can be converted to text
  • Palm Pilot graffiti can be converted to text
  • T9 converts numbers to text
  • Implicit input such as body gestures and facial
    expressions cannot be converted to text
  • Outside the scope of this project stage

7
Issues Involved Input Side
  • Understanding different input types
  • Inherent differences in information world
  • Processing method
  • Can it be generic?
  • How much customization is required?

8
What kind of output can we accommodate?
  • We must be able to control the application
    through another program so that our event handler
    can do the action requested by the input
  • Must take in at least one of the inputs that fit
    the input requirements

9
Issues Involved Output Side
  • Choosing the output node
  • Similar output applications could lead to the
    wrong decision.
  • What if two applications want the input? How
    should we define how many connections to make?
  • Different output formats
  • How much customization is required for each
    application?

10
First Generation Putting input and output
together
Appl. 1
Appl. 2
Appl. 3
Speech
Speech
Speech
11
Second Generation Switchbox Approach
Appl. 1
Appl. 2
Appl. 3
Switchbox
Speech
Speech
Speech
12
Third Generation Processing Input, Choosing
Output
Appl. 2
Appl. 1
Appl. 3
Events
Speech
13
Mutiple Output Demo
  • Speech Input
  • Multiple speech domain applications
  • Broadcast speech messages
  • Choose the best fit
  • Allows for controlling different layers
  • Galaxy speech processing system

14
Use Galaxy Components Individually
  • Utilizing server within Galaxy hub to access
    individual components

Recognizer
Natural Language Processor
Hub
15
Multiple Output Demo Diagram
Appl. 2
Appl. 1
Appl. 3
Speech
16
Future Work
  • Error tolerant system
  • Requires understanding of valid input and output
    types
  • Modular
  • Hard because each input modality requires
    different type of processing
  • Plug and play?
Write a Comment
User Comments (0)
About PowerShow.com