Vocoders 1 The Channel Vocoder (analyzer) : The channel vocoder - - PowerPoint PPT Presentation
Vocoders 1 The Channel Vocoder (analyzer) : The channel vocoder employs a bank of bandpass filters, Each having a bandwidth between 100 Hz and 300 Hz. Typically, 16-20 linear phase FIR filter are used. The output of each filter is
1
2
The channel vocoder employs a bank of
Each having a bandwidth between 100 Hz and 300
Hz.
Typically, 16-20 linear phase FIR filter are used.
The output of each filter is rectified and lowpass
The bandwidth of the lowpass filter is selected to
match the time variations in the characteristics of the vocal tract.
For measurement of the spectral magnitudes, a
3
Bandpass Filter A/D Converter Lowpass Filter A/D Converter Lowpass Filter Rectifier Rectifier Bandpass Filter Voicing detector Pitch detector Encoder
S(n) To Channel
4
16-20 linear-phase FIR filters Covering 0-4 kHz Each having a bandwidth between 100-
20-ms frames, or 50 Hz changing of
LPF bandwidth: 20-25 Hz Sampling rate of the output of the filters:
5
Bit rate:
1 bit for voicing detector 6 bits for pitch period For 16 channels, each coded with 3-4 bits,
Then the total bit rate is 2400-3200 bps Further reductions to 1200 bps can be
6
At the receiver the signal samples are passed
The outputs of the D/As are multiplied by the
The resulting signal are passed through
The outputs of the bandpass filters are summed
7
D/A Converter Decoder D/A Converter Voicing Information Pitch period Pulse generator Random Noise generator Bandpass Filter Bandpass Filter Switch ∑
Output speech From Channel
8
The phase vocoder is similar to the
However, instead of estimating the pitch,
By coding and transmitting the phase
9
(analyzer block diagram, kth channel)
n
k
cos n
k
sin n
k
cos
Lowpass Filter Encoder Lowpass Filter
Differentiator Differentiator
Decimator Decimator Compute Short-term Magnitude And Phase Derivative
To Channel S(n)
n
k
sin
n
k
cos
n ak
n bk
Short-term magnitude Short-term phase derivative
10
(synthesizer block diagram, kth channel)
n
k
cos
Interpolator Decoder ∑
From Channel
Cos Integrator Interpolator Sin
Decimated Short-term amplitude Decimated Short-term Phase derivative n
k
sin
11
LPF bandwidth: 50 Hz Demodulation separation: 100 Hz Number of filters: 25-30 Sampling rate of spectrum magnitude and phase
Spectral magnitude is coded using PCM or
Phase derivative is coded linearly using 2-3 bits The resulting bit rate is 7200 bps
12
The formant vocoder can be viewed as a
It is this information plus the pitch period
13
Example of formant:
(a) : The spectrogram of the utterance “day one”
showing the pitch and the harmonic structure of speech.
(b) : A zoomed spectrogram of the fundamental and
the second harmonic.
14
F3 F2 F1
Pitch
And
V/U
Decoder
F3 B3 F2 B2 F1 B1 V/U F0 Fk :The frequency of the kth formant Bk :The bandwidth of the kth formant Input Speech
15
F3 F2 F1
Excitation Signal
F3 B3 F2 B2 F1 B1 V/U F0 ∑
16
The objective of LP analysis is to estimate
Several methods have been devised for
Various LPC-type speech analysis and synthesis
17
This methods is called LPC-10 because of
LPC-10 partitions the speech into the 180
Pitch and voicing decision are determined
18
A General Discrete-Time Model For Speech Production
DT Impulse generator G(z) Glottal Filter
Uncorrelated
Noise generator H(z) Vocal tract Filter R(z) LP Filter Voiced Unvoiced Pitch Gain Gain V U
U(n) Voiced Volume velocity
s(n) Speech Signal
هحفص19 زا54
يطخ ييوگشيپ
هحفص20 زا54
يطخ ييوگشيپ
هحفص21 زا54
يطخ ييوگشيپ
m M m n m M m n
n e n s PG
1 2 1 2
] [ ] [ log 10
هحفص22 زا54
M=4 M=10
يطخ ييوگشيپ
هحفص23 زا54
M=2 M=10 M=54
يطخ ييوگشيپ
هحفص24 زا54
M=10 M=50
يطخ ييوگشيپ
هحفص25 زا54
يطخ ييوگشيپ
هحفص26 زا54
ردكو LPC10
هحفص27 زا54
LPC LPC
Bit Encoder
PCM LPC
ردكو LPC10
هحفص28 زا54
YMC
m N m n l] s[n]s[n R[l,m] 1
m N m n l n s n s m l MDF 1 ] [ ] [ ] , [
m N m n e N n s b n s 1 ], [ ] [ . ] [
هحفص29 زا54
MDF T=20,21,…,39,40,42,…,80,84 ,…,154
ردكو LPC10
هحفص30 زا54
LPC
RC
ردكو LPC10
هحفص31 زا54
يلصا لانگيس هدننك دك شخب
G يرپ اب هبرض راطقدو ماگ هرود رياري زيون يفداصت V/U هدش زتنس راتفگ
ردكو LPC10
هحفص32 زا54
AR
ردكو LPC10
33
Speech quality can be improved at the
One method is that the LPC model and
34
The speech is synthesized at the transmitter and
The residual error is quantized, coded, and
At the receiver the signal is synthesized by
The residual signal is low-pass filtered at 1000 Hz in the
analyzer to reduce bit rate
In the synthesizer, it is rectified and spectrum flattened
(using a HPF), the lowpass and highpass signals are summed and the resulting residual error signal is used to excite the LPC model.
RELP vocoder provides communication-quality speech
at about 9600 bps.
35
36
Buffer And window stLP analysis ∑ Encoder
LP Synthesis model
S(n) To Channel Excitation parameters
LP Parameters
f (n; m) e (n; m)
Residual error
m)} (i; a ˆ {
estimate pitch , P ˆ decision V/U, estimate gain , Θ ˆ
37
Buffer And window
S(n)
f (n; m)
Inverse Filter m) (z; A ˆ Lowpass Filter Decimator DFT Encoder
To Channel Prediction Residual
m) (n; stLP analysis
LP Parameters
m)} (i; a ˆ {
38
Decoder From Channel Buffer And Controller Interpolator Rectifier Highpass Filter
Residual
LP synthesizer
LP model Parameter updates
∑
Excitation
39
RELP needs to regenerate the high-
A crude approximation of the high frequencies
The multipulse LPC is a time domain
40
The information concerning the excitation sequence
includes:
the location of the pulses an overall scale factor corresponding to the largest pulse amplitude The pulse amplitudes relative to the overall scale factor
The scale factor is logarithmically quantized into 6 bits. The amplitudes are linearly quantized into 4 bits. The pulse locations are encoded using a differential
coding scheme.
The excitation parameters are updated every 5 msec. The LPC vocal-tract parameters and the pitch period are
updated every 20 msec.
The bit rate is 9600 bps.
41
) / ( ˆ ) ( ˆ ) ( ˆ ) / ( ˆ ) ( c z A z A z c z z W
A stored sequence from a Gaussian
The synthetic speech is compared with the
Residual error signal is weighted
42
(Analysis by synthesis method)
LP Synthesis filter Buffer And LP analysis Multipulse Excitation generator Error minimization Perceptual Weighting filter W(z)
∑ m) (n; f ˆ m) f(n; s(n) Input speech m) (n; m) (n;
W
+
Synthesis
(z) filterΘp
P ˆ
43
CELP is an analysis-by-synthesis method
The bit rate of the CELP is 4800 bps.
44
Gaussian Excitation codebook
Pitch Synthesis filter
Spectral Envelope (LP) Synthesis filter
∑
Perceptual Weighting Filter W(z)
Computer Energy (square and sum)
Buffer and LP analysis
Side information Gain
LP parameters Speech samples
Index of Excitation sequence
45
This weighted error is squared and
By performing an exhaustive search
46
The gain factor for scaling the excitation
47
From Channel decoder
Buffer And controller Gaussian Excitation codebook Pitch Synthesis filter LP Synthesis filter LP parameters, gain and pitch estimate updates
48
Cascade of two all-pole filter with coefficients
First filter is a long-delay pitch filter used to
This filter has this form
p p p
49
Parameters of the filter can be determined
Second filter is a short-delay all-pole
50
sampling frequency is 8khz
We have 40 samples per 5-msec The excitation sequence consist of 40
51
A codebook of 1024 sequences gives
For such codebook size ,we require
Hence the bit rate is reduced by a factor
The transmission of pitch predictor
52
CELP has been used to achieve toll-
Although other types of vocoders
53
The one way delay is of the order of 20-40
With modification of CELP, it is possible to
Low-delay CELP is achieved by using a
54
Excitation Vector quantizer codebook
LP (high-order) Synthesis filter
∑
Perceptual Weighting Filter W(z)
Error minimization
Buffer and window
Input Speech
+
(n; m) (n;
W
m) f(n; s(n)
Gain Gain adaptation Predictor adaptation
m) (n; f ˆ
55
Pitch predictor used in the conventional
In order to compensate for the loss in pitch
56
LPC coefficients are updated more
5-sample excitation vector corresponds to
57
The logarithm of the excitation gain is
The coefficients of the logarithmic-gain
58
The perceptual weighting filter is also 10th
The excitation codebook in the low-delay
10-bit excitation codebook is employed
59
The VSELP coder and decoder basically differ in
In the next block diagram of the VSELP, there
One excitation is obtained from the pitch period
The other two excitation sources are obtained
60
1
Long-term Filter state Codebook 1 Codebook 2 ∑ Pitch synthesis filter Spectral post filter
Spectral envelop (LP) synthesis filter Synthetic Speech
2
61
LPC synthesis filter is implemented as a
Coefficients are updated in each 5-ms
Excitation parameters are also updated
62
128 codewords in each of the two
codewords are constructed from two sets
The long-term filter state is also a
63
In each 5-msec frame, the codewords from
The filtered codeword is used to update
64
Thus the update occurs by appending the
The oldest sample in the history array is
The result is that the long-term state
65
The three excitation sequences are
Each codebook search attempts to find the
Once the codewords have been selected
66
Joint gain optimization is sequentially
These parameters are vector quantized to
67
The bit rate of the VSELP is about 8000 bps.
Bit allocations for 8000-bps VSELP
Parameters Bits/5-ms Frame Bits/20ms
10 LPC coefficients - 38 Average speech energy - 5
Excitation codewords from two VSELP codebooks 14 56
Gain parameters 8 32 Lag of pitch filter 7 28 Total
29 159
68
Finally, an adaptive spectral post filter is
69
Speech Codec Male Speaker Female Speaker Music Original Speech/Music (16-bit sampled at 8KHz) FS-1015 (LPC-10e 2.4 kb/s) FS-1016(CELP 4.8 kb/s) IS-54 ( VSELP 7.95 kb/s) G.721 (32 kb/s ADPCM)
Standard Voice Algorithms G.711
The most widely used digital representation of voice signals is that of
the G.711 or PCM (Pulse Code Modulation)
This codec represents a 4 kHz band limited voice signal sampled at 8
kHz using 8 bits per sample A-law or m-law coding. G.726
The protocol for the G.726 codec requires a 64 kbps A-Law or m-law
PCM signal to be encoded into four different bit rate options ranging from 2 bits per sample to 5 bits per sample
The algorithm is based on Adaptive Differential Pulse Code Modulation
(ADPCM) and is based on 1 sample backward prediction scheme.
70
G.728
The G.728 algorithm compresses PCM codec voice signals to a bit rate of 16 kbps.
This algorithm is based on a strong backward prediction scheme and is by far considered as one
G.729
For compression of voice signals at 8 kbps the G.729 algorithm offers toll quality with built in algorithmic delays of less than 15 msec
Additional features described in the G.729 Annex ensure VAD1 and Comfort Noise Generation functionalities to enhance the quality and reduce the overall bit rate
G.723.1
The most widely used algorithm for band limited channels, such as VoIP and video conferencing, is that of G.723.1
The algorithm has two operating bit rates of 6.3 kbps and 5.3 kbps
Although the delay is not as low as that of the other ITU standards its quality is near toll quality for the given low bit rates, making it very efficient in bit usage.
71
GSM2—AMR
The latest GSM standard is the multi rate Adaptive Code Excited Linear Prediction that provides compression in the range of 4.75 to 12.2 kbps
In total the codec provides 12 bit rates that cover the half rate to full rate channel capacity.
GSM—FR
The first digital codec used in a mobile environment is the GSM Full Rate vocoder
The codec compresses 13 bit PCM sample signals to a rate of 13 kbps
The algorithm is based on a very simple Regular Pulse Excited – Linear Prediction Coding technique.
GSM—HR
To increase capacity, the GSM committee decided on a lower bit rate of 5.6 kbps for the voice channel
The algorithm is based on the Vector Sum Excited Linear Predictive (VSELP) and is computationally as complex as other low bit rate algorithms.
72