SLIDE 1
Filter Banks
SPEECH RECOGNITION 40833
1
Spectral Analysis Models
2
- (a) Pattern Recognition
- (b) Acoustic phonetic
approaches to speech recognition
Filter Banks SPEECH RECOGNITION 40833 1 2 Spectral Analysis - - PowerPoint PPT Presentation
Filter Banks SPEECH RECOGNITION 40833 1 2 Spectral Analysis - - PowerPoint PPT Presentation
Filter Banks SPEECH RECOGNITION 40833 1 2 Spectral Analysis Models (a) Pattern Recognition (b) Acoustic phonetic approaches to speech recognition 3 Spectral Analysis Models LPC analysis model 4 THE BANK-OF-FILTERS FRONT- END
SLIDE 2
SLIDE 3
Spectral Analysis Models
3
- LPC analysis model
SLIDE 4
THE BANK-OF-FILTERS FRONT- END PROCESSOR
4
- Complete bank-of-filter analysis model
SLIDE 5
THE BANK-OF-FILTERS FRONT- END PROCESSOR
5
SLIDE 6
THE BANK-OF-FILTERS FRONT- END PROCESSOR
6
Typical waveforms and spectra for analysis of a pure sinusoid in the filter-bank model
SLIDE 7
THE BANK-OF-FILTERS FRONT- END PROCESSOR
7
Typical waveforms and spectra of a voice speech signal in the bank-of-filters analysis model
SLIDE 8
THE BANK-OF-FILTERS FRONT- END PROCESSOR
8
Ideal (a) and realistic (b) set of filter responses of a Q-channel filter bank covering the frequency range Fs/N to (Q+1/2)Fs/N
SLIDE 9
Types of Filter Bank Used for Speech Recognition
N F b N/2 Q Q i 1 i, N F f
s i s i
9
SLIDE 10
Non-uniform Filter Banks
1 i 1 j 1 i j 1 i 1 i i 1
, 2 ) b (b b f f Q i 2 , b α b c b
10
SLIDE 11
Nonuniform Filter Banks
1600Hz b Hz, 2400 f : 4 Filter Hz 800 b Hz, 1200 f : 3 Filter Hz 400 b Hz, 600 f : 2 Filter 200Hz b Hz, 300 f : 1 Filter
4 4 3 3 2 2 1 1
11
SLIDE 12
Types of Filter Bank Used for Speech Recognition
12
Ideal specification of a 4-channel octave band-filter bank (a), a 12-channel third-octave band filter bank (b), and a 7-channel critical band scale filter bank
SLIDE 13
Types of Filter Bank Used for Speech Recognition
13
The variation of bandwidth with frequency for the perceptually based critical band scale Ideal specification of a 4-channel octave band-filter bank (c) covering the telephone bandwidth range (200-3200 HZ)
SLIDE 14
Implementations of Filter Banks
14
- Instead of direct convolution, which is computationally expensive, we
assume each bandpass filter impulse response to be represented by:
- hi(n), i-th bandpass filter impulse response, is represented by a fixed
lowpass window, w(n), modulated by complex exponential
- Where w(n) is a fixed lowpass window representing the
i
j n
e
SLIDE 15
Implementations of Filter Banks
15
The signals s(m) and w(n-m) used in evaluation of the short-time Fourier transform
SLIDE 16
Frequency Domain Interpretation of the Short-Time Fourier Transform
16
Short-time Fourier transform using a long (500 points or 50 msec) Hamming window on a section of voiced speech
FREQUENCY SAMPLE VALUE LOG MAGNITUDE (dB)
SLIDE 17
Frequency Domain Interpretation of the Short-Time Fourier Transform
17
Short-time Fourier transform using a short (50 points or 5 msec) hamming window on a section of voiced speech
FREQUENCY SAMPLE VALUE LOG MAGNITUDE (dB)
SLIDE 18
Frequency Domain Interpretation of the Short-Time Fourier Transform
18
FREQUENCY SAMPLE VALUE LOG MAGNITUDE (dB)
Short-time Fourier transform using a long (500 points or 50 msec) hamming window on a section of unvoiced speech
SLIDE 19
Frequency Domain Interpretation of the Short-Time Fourier Transform
19
FREQUENCY SAMPLE VALUE LOG MAGNITUDE (dB)
Short-time Fourier transform using a short (50 points or 5 msec) hamming window on a section of unvoiced speech
SLIDE 20
Linear Filter Interpretation of the STFT
20
) (
~ n
s
) (n s
) (n w
i
j
e
) (
1
j n e
S
SLIDE 21
FFT Implementation of a Uniform Filter Bank
21
FFT implementation of a uniform filter bank
SLIDE 22
Direct implementation of an arbitrary filter bank
22
) (n s
) (
1 n
h ) (n X Q ) (
2 n
h
) (n hQ
) (
1 n
X
) (
2 n
X
SLIDE 23
Nonuniform FIR Filter Bank Implementations
23
Two arbitrary nonuniform filter-bank filter specifications consisting
- f eighter 3 bands (part a) or 7 bands (part b).
SLIDE 24
Tree Structure Realizations of Nonuniform Filter Banks
24
SLIDE 25
Practical Examples of Speech-Recognition Filter Banks
25
FREQUENCY (kHz) TIME IN SAMPLES VALUE MAGNITUDE (dB)
SLIDE 26
Practical Examples of Speech-Recognition Filter Banks
26
Window sequence, w(n), (part a), the individual filter response (part b), and the composite response (part c) of a Q = 15 channel, uniform filter bank, designed singal 101-point Kaiser window smoothed lowpass window (after Dautrich et al).
FREQUENCY (kHz) MAGNITUDE (dB)
SLIDE 27
Practical Examples of Speech-Recognition Filter Banks
27
FREQUENCY (kHz) TIME IN SAMPLES VALUE MAGNITUDE (dB)
SLIDE 28
Practical Examples of Speech-Recognition Filter Banks
28
Window sequence, w(n), (part a), the individual filter response (part b), and the composite response (part c) of a Q = 15 channel, uniform filter bank, designed singal 101-point Kaiser window directly as the lowpass window (after Dautrich etal).
FREQUENCY (kHz) MAGNITUDE (dB)
SLIDE 29
Generalizations of Filter-Bank Analyzer
29
SLIDE 30
30
Generalizations of Filter-Bank Analyzer
SLIDE 31
31
Generalizations of Filter-Bank Analyzer
SLIDE 32
32