applications and derivation of linear predictive coding

54
linear predictive coding (LPC) an application driven approach adapted from guest lecture for mobile application development for sensing and control, EE596 Friday, August 30, 13

Upload: eric-larson

Post on 25-Dec-2014

332 views

Category:

Technology


0 download

DESCRIPTION

 

TRANSCRIPT

Page 1: Applications and Derivation of Linear Predictive Coding

linear predictive coding (LPC)an application driven approach

adapted from guest lecture for mobile application development for sensing and control, EE596

Friday, August 30, 13

Page 2: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

Data

Derivative

Friday, August 30, 13

Page 3: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

Data

Derivative

Friday, August 30, 13

Page 4: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

Data

Derivative

derivative[n] = y[n]-y[n-1]

Friday, August 30, 13

Page 5: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

Data

Derivative

derivative[n] = y[n]-y[n-1]

Friday, August 30, 13

Page 6: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

Data

Derivative

derivative[n] = y[n]-y[n-1]

Friday, August 30, 13

Page 7: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

Data

Derivative

cos(x)

derivative[n] = y[n]-y[n-1]

Friday, August 30, 13

Page 8: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

Data

Derivative

cos(x)

derivative[n] = y[n]-y[n-1]

-sin(x)

Friday, August 30, 13

Page 9: Applications and Derivation of Linear Predictive Coding

the tradeoff of parametric modeling

Friday, August 30, 13

Page 10: Applications and Derivation of Linear Predictive Coding

the tradeoff of parametric modeling

- need to fit a model to the data

Friday, August 30, 13

Page 11: Applications and Derivation of Linear Predictive Coding

the tradeoff of parametric modeling

- need to fit a model to the data

+ (might be) easier to manipulate model

Friday, August 30, 13

Page 12: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

two signals = 1500 Hz and 5500 Hz two signals = 1500 Hz and 5500 Hz

Magn

itude

freq (kHz) freq (kHz)

Magn

itude

Friday, August 30, 13

Page 13: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

two signals = 1500 Hz and 5500 Hz

FFT, array

two signals = 1500 Hz and 5500 Hz

Magn

itude

freq (kHz) freq (kHz)

Magn

itude

Friday, August 30, 13

Page 14: Applications and Derivation of Linear Predictive Coding

non-parametric parametricUse Data or Transform Fit Data to a Model

two signals = 1500 Hz and 5500 Hz

FFT, array

two signals = 1500 Hz and 5500 Hz

Magn

itude

freq (kHz)

LPC polynomialfreq (kHz)

Magn

itude

Friday, August 30, 13

Page 15: Applications and Derivation of Linear Predictive Coding

what model should we fit to?

Friday, August 30, 13

Page 16: Applications and Derivation of Linear Predictive Coding

what model should we fit to?

a filter with feedback

Friday, August 30, 13

Page 17: Applications and Derivation of Linear Predictive Coding

what model should we fit to?

a filter with feedback

Friday, August 30, 13

Page 18: Applications and Derivation of Linear Predictive Coding

what model should we fit to?

a filter with feedback

Friday, August 30, 13

Page 19: Applications and Derivation of Linear Predictive Coding

feedback filters are system models

Friday, August 30, 13

Page 20: Applications and Derivation of Linear Predictive Coding

feedback filters are system models

Friday, August 30, 13

Page 21: Applications and Derivation of Linear Predictive Coding

feedback filters are system models

Friday, August 30, 13

Page 22: Applications and Derivation of Linear Predictive Coding

feedback filtering

a

Friday, August 30, 13

Page 23: Applications and Derivation of Linear Predictive Coding

feedback filtering

want to estimate a

a

Friday, August 30, 13

Page 24: Applications and Derivation of Linear Predictive Coding

feedback filtering

what can we represent with this equation?

Friday, August 30, 13

Page 25: Applications and Derivation of Linear Predictive Coding

ak

k k k

feedback filtering

what can we represent with this equation?

3210-1-2-3

1 3 5 7 9 11 13

3210-1-2-3

1 3 5 7 9 11 13

3210-1-2-3

1 3 5 7 9 11 13

piano marimba violin

Friday, August 30, 13

Page 26: Applications and Derivation of Linear Predictive Coding

feedback filter equation in frequency

Friday, August 30, 13

Page 27: Applications and Derivation of Linear Predictive Coding

feedback filter equation in frequency

Y (z) =E(z)

1�Pp

k=1 akz�k

z = ej!

Friday, August 30, 13

Page 28: Applications and Derivation of Linear Predictive Coding

is this a good model for frequency analysis?

Y (z) =1

1�Pp

k=1 akz�k

E(z)

Y (z) =1Qp

k=1(1� rkz�1)E(z)

Friday, August 30, 13

Page 29: Applications and Derivation of Linear Predictive Coding

is this a good model for frequency analysis?

resonant frequency = complex angle of rootresonance bandwidth = related to magnitude of root

Y (z) =1

1�Pp

k=1 akz�k

E(z)

Y (z) =1Qp

k=1(1� rkz�1)E(z)

Friday, August 30, 13

Page 30: Applications and Derivation of Linear Predictive Coding

examples

Y (z) =1Qp

k=1(1� rkz�1)E(z)

Friday, August 30, 13

Page 31: Applications and Derivation of Linear Predictive Coding

another interpretation, vocal tract

sourcefilter

Y (z) =1

1�Pp

k=1 akz�k

E(z)

Friday, August 30, 13

Page 32: Applications and Derivation of Linear Predictive Coding

another interpretation, vocal tract

sourcefilter

Y (z) =1

1�Pp

k=1 akz�k

E(z)

Friday, August 30, 13

Page 33: Applications and Derivation of Linear Predictive Coding

another interpretation, vocal tract

sourcefilter

Y (z) =1

1�Pp

k=1 akz�k

E(z)

Friday, August 30, 13

Page 34: Applications and Derivation of Linear Predictive Coding

another interpretation, prediction

Friday, August 30, 13

Page 35: Applications and Derivation of Linear Predictive Coding

another interpretation, prediction

Friday, August 30, 13

Page 36: Applications and Derivation of Linear Predictive Coding

17

Friday, August 30, 13

Page 37: Applications and Derivation of Linear Predictive Coding

18

Friday, August 30, 13

Page 38: Applications and Derivation of Linear Predictive Coding

18

Friday, August 30, 13

Page 39: Applications and Derivation of Linear Predictive Coding

18

Friday, August 30, 13

Page 40: Applications and Derivation of Linear Predictive Coding

summary of interpretations

Friday, August 30, 13

Page 41: Applications and Derivation of Linear Predictive Coding

summary of interpretations

Spectral Estimation == Auto Regressive

Friday, August 30, 13

Page 42: Applications and Derivation of Linear Predictive Coding

summary of interpretations

Spectral Estimation == Auto RegressiveForecasting == Linear Prediction

Friday, August 30, 13

Page 43: Applications and Derivation of Linear Predictive Coding

summary of interpretations

Spectral Estimation == Auto RegressiveForecasting == Linear Prediction

Vocal Tract Model == Source/Filter

Friday, August 30, 13

Page 44: Applications and Derivation of Linear Predictive Coding

common applications

Friday, August 30, 13

Page 45: Applications and Derivation of Linear Predictive Coding

common applications

Speech Vocoders

Friday, August 30, 13

Page 46: Applications and Derivation of Linear Predictive Coding

common applications

Speech VocodersSpectral Analysis

Friday, August 30, 13

Page 47: Applications and Derivation of Linear Predictive Coding

common applications

Speech VocodersSpectral AnalysisPitch Estimation

Friday, August 30, 13

Page 48: Applications and Derivation of Linear Predictive Coding

common applications

Speech VocodersSpectral AnalysisPitch Estimation Voice Changers

Friday, August 30, 13

Page 49: Applications and Derivation of Linear Predictive Coding

common applications

Speech VocodersSpectral AnalysisPitch Estimation Voice Changers

Friday, August 30, 13

Page 50: Applications and Derivation of Linear Predictive Coding

common applications

Speech VocodersSpectral AnalysisPitch Estimation Voice Changers Analysis/Synthesis of Instrument Sounds

Friday, August 30, 13

Page 51: Applications and Derivation of Linear Predictive Coding

common applications

Speech VocodersSpectral AnalysisPitch Estimation Voice Changers Analysis/Synthesis of Instrument Sounds Voice Box

Friday, August 30, 13

Page 52: Applications and Derivation of Linear Predictive Coding

common applications

Speech VocodersSpectral AnalysisPitch Estimation Voice Changers Analysis/Synthesis of Instrument Sounds Voice Box

Compression (i.e., mpeg4, CELP)

Friday, August 30, 13

Page 53: Applications and Derivation of Linear Predictive Coding

common applications

Speech VocodersSpectral AnalysisPitch Estimation Voice Changers Analysis/Synthesis of Instrument Sounds Voice Box

Compression (i.e., mpeg4, CELP)

My research– medical sensing from a microphone

Friday, August 30, 13

Page 54: Applications and Derivation of Linear Predictive Coding

questions?Topics Related to LPC and Further Reading:

LPC10, Ultra Low Bit Rate Voice CodingCode Excited Linear PredictionLevinson-Durbin RecursionBurg’s MethodLP Cepstral CoefficientsThe Talking OrchestraSpiroSmart, the mobile phone spirometer

eclarson.com [email protected]@ec_larson

electrical engineering

computerscience

Friday, August 30, 13