Frequency Domain Coding of Speech - PowerPoint PPT Presentation

1 / 65
About This Presentation
Title:

Frequency Domain Coding of Speech

Description:

The Short-Time Fourier Transform. The Short-Time Discrete ... Spectrogram. R. 2R. 3R. 4R. Sampled STDFT. Analysis: Synthesis: In what condition we will have ... – PowerPoint PPT presentation

Number of Views:314
Avg rating:3.0/5.0
Slides: 66
Provided by: aimm02Cs
Category:

less

Transcript and Presenter's Notes

Title: Frequency Domain Coding of Speech


1
Frequency Domain Coding of Speech
  • ??????

2
Content
  • Introduction
  • The Short-Time Fourier Transform
  • The Short-Time Discrete Fourier Transform
  • Wide-Band Analysis/Synthesis
  • Sub-Band Coding

3
Frequency Domain Coding of Speech
  • Introduction

4
Speech Coders
  • Waveform Coders
  • Attempt to reproducing the original waveform
    according to some fidelity criteria
  • Performance successful at producing good
    quality, robust speech.
  • Vocoders
  • Correlated with speech production model.
  • Performance more fragile and more model
    dependent.
  • Lower bit rate

5
Frequency-Domain Coders
  • Sub-band coder (SCB).
  • Adaptive Transform Coding (ATC).
  • Multi-band Excited Vocoder (MBEV).
  • Noise Shaping in Speech Coders.

6
Classification of Speech Coders
7
Frequency Domain Coding of Speech
  • The Short-Time Fourier Transform

8
Definition of STFT
Interpretations
  • Filter Bank Interpretation
  • Block Transform Interpretation

9
Filter Bank Interpretation
f (m)
Analysis Filter
? is fixed at ?0.
10
Filter Bank Interpretation
11
Filter Bank Interpretation
Modulation
12
Filter Bank Interpretation
Modulation
Lowpass Filter
13
Filter Bank Interpretation
Modulated Subband signals
14
Block Transform Interpretation
n is fixed at n0.
Analysis Window
FT of Windowed Data
Windowed Data
15
Block Transform Interpretation
n is fixed at n0.
. . .
16
Analysis/Synthesis Equations
Analysis
Synthesis
17
Analysis/Synthesis Equations
Analysis
Synthesis
Replace r with nr
18
Analysis/Synthesis Equations
Analysis
Synthesis
19
Analysis/Synthesis Equations
Analysis
Synthesis
More general,
20
Examples
21
Examples
h(0)x(n)
22
Examples
23
Frequency Domain Coding of Speech
  • The Short-Time Discrete Fourier Transform

24
Definition of STDFT
Analysis
Synthesis
25
Synthesis
1
26
Synthesis
We need only one period.
Therefore, the condition is respecified as
27
Implementation Consideration
28
Sampling
R
2R
3R
4R
29
Sampled STDFT
Analysis
Synthesis
30
Sampled STDFT
Analysis
Synthesis
31
Frequency Domain Coding of Speech
  • Wide-Band
  • Analysis/Synthesis

32
Short-Time Synthesis --- Filter Bank Summation
STFT
Lowpass Filter
33
Short-Time Synthesis --- Filter Bank Summation
STFT
34
Short-Time Synthesis --- Filter Bank Summation
Lowpass filter
Bandpass filter
35
Short-Time Synthesis --- Filter Bank Summation
Lowpass representation of for the signal
in a band centered at ?k.
36
Short-Time Synthesis --- Filter Bank Summation
Encoding one band
Decoding one band
37
Short-Time Synthesis --- Filter Bank Summation
. . .
Analysis
Synthesis
38
Short-Time Synthesis --- Filter Bank Summation
. . .
Analysis
Synthesis
39
Short-Time Synthesis --- Filter Bank Summation
h1(n)
. . .
Analysis
Synthesis
40
Equal Spaced Ideal Filters
N 6
41
Equal Spaced Ideal Filters
What condition should be satisfied so that
y(n)x(n)?
42
Equal Spaced Ideal Filters
Time-Aliased version of h(n)
Equal spaced sampling of H(ej?)
Inverse discrete FT of H(ej?)
43
Equal Spaced Ideal Filters
In case that N ? L,
Consider FIR, i.e., h(n) is of duration of L
samples.
44
Equal Spaced Ideal Filters
45
Equal Spaced Ideal Filters
x(n) can always be Reconstructed if N ? L,
46
Equal Spaced Ideal Filters
Does x(n) can still be reconstructed if NltL?
If affirmative, what condition should be
satisfied?
x(n) can always be Reconstructed if N ? L,
47
Equal Spaced Ideal Filters
p(n)
48
Equal Spaced Ideal Filters
Signal can be reconstructed If it equals to
?(n?m).
p(n)
49
Typical Sequences of h(n)
Ideal lowpass filter with cutoff at ?/N.
50
Typical Sequences of h(n)
N ? L
51
Typical Sequences of h(n)
A causal FIR lowpass filter
52
Typical Sequences of h(n)
53
Filter Back Implementation for a Single Channel
Analysis
Synthesis
54
Filter Back Implementation for a Single Channel
Decimator
Interpolator
Analysis
Synthesis
55
Filter Back Implementation for a Single Channel
Depends on the bandwidth of h(n).
Decimator
Interpolator
R?
Analysis
Synthesis
56
Frequency Domain Coding of Speech
  • Sub-Band Coding

57
Filter Bank Implementation(Direct Implementation)
Complex Channels
R2??B
Bandwidth B/2
Analysis
Synthesis
58
Filter Bank Implementation(Practical
Implementation)
59
Filter Bank Implementation(Practical
Implementation)
60
Filter Bank Implementation(Practical
Implementation)
61
Filter Bank Implementation(Practical
Implementation)
Why?
62
Filter Bank Implementation(Practical
Implementation)
63
Filter Bank Implementation(Practical
Implementation)
64
Filter Bank Implementation(Practical
Implementation)
65
Filter Bank Implementation(Practical
Implementation)
ADPCM CODEC
Filter Bank Analysis
Sub-Band Coder Modification
Filter Bank Synthesis
Write a Comment
User Comments (0)
About PowerShow.com