sounds of silence the challenge for ai b.yegnanarayana speech and vision lab dept. of cs&e, iit...

20
Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Upload: martin-mathews

Post on 04-Jan-2016

218 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Sounds of Silence The Challenge for AI

B.YegnanarayanaSpeech and Vision Lab

Dept. of CS&E, IIT Madras

Page 2: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

The urge is to make machines more intelligentBut in the process we are doing the opposite

Why? Because we are only storing and manipulating data

What is INTELLIGENCE? It is not simply manipulation of data

Intelligence of human beingsCapture, associate and retrieve patterns

ExamplesSignatures, face recognition, video

Goal of Artificial Intelligence

Next >< Prev

Page 3: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Representation of 1D, 2D, 3D data

ContrastReading vs writingListening vs speakingLooking vs sketchingWatching vs doingRecognition vs synthesis

Key is learning Development of motor control is slow

Intelligent activityInvolves linking key features/concepts/ideas

Why AI is difficult: Some examples

Next >< Prev

Page 4: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Intelligence vs Information

Creating an environment for intelligent activityCurrent methods do exactly the opposite

We present more data more frequentlyNo scope for acquiring implicit pattern behavior

Confusion between knowledge and informationKnowledge society or ignorance society

Filling up mind with data is like filling up silence Intelligence is in capturing the sounds of silence

Next >< Prev

Page 5: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Sounds of SilenceSignificance of silence

In cartoons and string of characters

Examples of silence in speech soundsA sufficient cueA necessary cue

Illustrations from continuous speechWaveform, residual and impulsesDifferent speakersDifferent languages

Perception of Sounds of SilenceHuman ability and machine's inability - Why?

Next >< Prev

Page 6: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Architectural MismatchMachines take mostly silence and may ignore signal

Machine Human

Representation Pixels/samples(mostly silence data)

Symbols andinterrelations

(ignores silence)Processor Single Multiple

(neurons)

Processing Sequential(local)

Parallel anddistributed

(local and global)

Next >< Prev

Page 7: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Nature of AI ProblemsData I nformation Knowledge I ntelligence

NaturalLanguage

String ofCharacters

Words/Sound Units

Rules(Syntax)

Message

SpeechSequence of

SamplesFormants &

Pitch

Intonation &Duration

(languageconstraints)

Message

ImageArray ofPixels

Objects &Interrelation

Rules to form picture

Message

Decisionmaking

Next >< Prev

Page 8: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Illustrations from Speech

• Nature of speech production and perception

• Challenges in speech recognition, synthesis, and speaker recognition

• Why they are difficult for machine and easy for us?

• Due to our ability to capture sounds of silence

Next >< Prev

Page 9: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Characteristics of Human Problem Solving

• Computing sounds of silence

• Essentially pattern processing instead of data processing

• Integration of local and global patterns

• Delayed decisions

• Nonuniqueness of solutions

Next >< Prev

Page 10: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Architectural Features of Possible New Models

Need to move from

• Deterministic computation to decision logic

• Sequential processing to PDP

• Set of equations to set of inequalities

• Problem solving to learning

• Data processing to multidimensional pattern processing

Next >< Prev

Page 11: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Conclusions• Powerful computers need not solve intelligent

problems

• Finer sampling need not result in good solutions

• Two interesting problems: Video processing and dictation machine

• The challenge is computing the sounds of silence

• Unless we watch, the technology may destroy itself by exposing its limitations.

• Dont forget that it is always the human being is the reference not the machine for intellectual abilities.

Next >< Prev

Page 12: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Thank you

Next >< Prev

Page 13: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Back

Next >< Prev

Page 14: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Some Illustrations of “Sounds of Silence”

30 msslit

150 ms split

600 ms s_lit

10 ms

100 ms

sha

shka

Silence:

A sufficient cue for stop

consonant perception

Silence:

A necessary cue for stop

consonant perception

BackNext >< Prev

Page 15: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Signal

Residual

Instants

Some Illustrations of “Sounds of Silence”

Back

More examples : Signal, residual and instants

Some more examples : Signal, residual and instants

Next >< Prev

Page 16: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Speech Production Mechanism

BackNext >< Prev

Page 17: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Still Frame Video Sequence

Less noise

More noise

Next >< Prev

Page 18: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Still Frame Video Sequence

Less noise

Morenoise

Next >< Prev

Page 19: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

Back

< Prev

Page 20: Sounds of Silence The Challenge for AI B.Yegnanarayana Speech and Vision Lab Dept. of CS&E, IIT Madras

< Prev