Upload
abhiram-yanamandra
View
231
Download
0
Embed Size (px)
Citation preview
8/3/2019 ch5.3 (Vocoders)
1/23
1
Vocoders
8/3/2019 ch5.3 (Vocoders)
2/23
2
The Channel Vocoder (analyzer):
The channel vocoder employs a bank ofbandpass 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 rectified and lowpassfiltered. The bandwidth of the lowpass filter is selected to
match the time variations in the characteristics of thevocal tract.
For measurement of the spectral magnitudes, avoicing detector and a pitch estimator areincluded in the speech analysis.
8/3/2019 ch5.3 (Vocoders)
3/23
3
The Channel Vocoder (analyzer block diagram):
BandpassFilter
A/DConverter
LowpassFilter
A/DConverter
LowpassFilter
Rectifier
Rectifier
BandpassFilter
Voicingdetector
Pitchdetector
Encoder
S(n)To
Channel
8/3/2019 ch5.3 (Vocoders)
4/23
4
The Channel Vocoder (synthesizer):
At the receiver the signal samples are passedthrough D/A converters.
The outputs of the D/As are multiplied by the
voiced or unvoiced signal sources.
The resulting signal are passed throughbandpass filters.
The outputs of the bandpass filters are summedto form the synthesized speech signal.
8/3/2019 ch5.3 (Vocoders)
5/23
5
The Channel Vocoder (synthesizer block diagram):
D/AConverter
Decoder
D/AConverter
VoicingInformation
Pitchperiod
Pulsegenerator
RandomNoise
generator
BandpassFilter
BandpassFilter
Switch
Output
speech
From
Channel
8/3/2019 ch5.3 (Vocoders)
6/23
6
The Phase Vocoder :
The phase vocoder is similar to thechannel vocoder.
However, instead of estimating the pitch,the phase vocoder estimates the phasederivative at the output of each filter.
By coding and transmitting the phasederivative, this vocoder destroys the phaseinformation .
8/3/2019 ch5.3 (Vocoders)
7/23
7
The Phase Vocoder (analyzer block diagram):
nk
cos nk
sin
nk
cos
LowpassFilter
Encoder
LowpassFilter
Differentiator
Differentiator
Decimator
Decimator
ComputeShort-termMagnitude
AndPhase
Derivative
To
Channel
S(n)
nk
sin
nk
cos
nak
nbk
Short-termmagnitude
Short-term phasederivative
8/3/2019 ch5.3 (Vocoders)
8/23
8
The Phase Vocoder(synthesizer block diagram, kth channel):
nkcos
Interpolator
Decoder
From
ChannelCos
Integrator
InterpolatorSin
DecimateShort-term
amplitude
Decimate
Short-term
Phase
derivative
nk
sin
8/3/2019 ch5.3 (Vocoders)
9/23
9
The Formant Vocoder :
The formant vocoder can be viewed as atype of channel vocoder that estimate thefirst three or four formants in a segment of
speech.
It is this information plus the pitch period
that is encoded and transmitted to thereceiver.
8/3/2019 ch5.3 (Vocoders)
10/23
10
The Formant Vocoder :
Example of formant: (a) : The spectrogram of the utterance day one
showing the pitch and the harmonic structure ofspeech.
(b) : A zoomed spectrogram of the fundamental andthe second harmonic.
(a) (b)
8/3/2019 ch5.3 (Vocoders)
11/23
11
The Formant Vocoder (analyzer block diagram):
F3
F2
F1
PitchAnd
V/UDecoder
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
8/3/2019 ch5.3 (Vocoders)
12/23
12
The Formant Vocoder (synthesizer block diagram):
F3
F2
F1
ExcitationSignal
F3
B3
F2
B2
F1
B1
V/U
F0
8/3/2019 ch5.3 (Vocoders)
13/23
8/3/2019 ch5.3 (Vocoders)
14/23
14
LPC 10 :
This methods is called LPC-10 because of10 coefficient are typically employed.
LPC-10 partitions the speech into the 180sample frame.
Pitch and voicing decision are determinedby using the AMDF and zero crossingmeasures.
8/3/2019 ch5.3 (Vocoders)
15/23
15
Residual Excited LP Vocoder :
Speech quality in speech quality can beimproved at the expense of a higher bitrate by computing and transmitting a
residual error, as done in the case of DPCM.
One method is that the LPC model and
excitation parameters are estimated froma frame of speech.
8/3/2019 ch5.3 (Vocoders)
16/23
16
Residual Excited LP Vocoder :
The speech is synthesized at the transmitter andsubtracted from the original speech signal toform the residual error.
The residual error is quantized, coded, andtransmitted to the receiver
At the receiver the signal is synthesized byadding the residual error to the signal generatedfrom the model.
8/3/2019 ch5.3 (Vocoders)
17/23
17
RELP Block Diagram :
BufferAnd
window
LPanalysis
Encoder
LPSynthesis
model
S(n)
To
ChannelExcitation
parameters
LP
Parameters
8/3/2019 ch5.3 (Vocoders)
18/23
18
Code Excited LP :
CELP is an analysis-by-synthesis methodin which the excitation sequence isselected from a codebook of zero-mean
Gaussian sequence.
The bit rate of the CELP is 4800 bps.
8/3/2019 ch5.3 (Vocoders)
19/23
19
CELP (analysis-by-synthesis coder) :
GaussianExcitationcodebook
PitchSynthesis
filter
Spectral
Envelope(LP)
Synthesis filter
PerceptualWeighting
Filter W(z)
ComputeEnergy of Error
(square and sum)
Buffer andLP
analysis
Side
informationGain
LP
parameters
Speech samples
Index of
Excitation
sequence
+
-
8/3/2019 ch5.3 (Vocoders)
20/23
20
CELP (synthesizer) :
From
Channeldecoder
BufferAnd
controller
GaussianExcitationcodebook
PitchSynthesis
filter
LPSynthesis
filter
LP parameters,
gain and pitchestimate
updates
8/3/2019 ch5.3 (Vocoders)
21/23
21
Vector Sum Excited LP :
The VSELP coder and decoder basically differ inmethod by which the excitation sequence isformed.
In next block diagram of the VSELP, there arethree excitation source.
One excitation is obtained from the pitch periodstate.
The other two excitation source are obtainedfrom two codebook.
8/3/2019 ch5.3 (Vocoders)
22/23
22
Vector Sum Excited LP :
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 56Gain parameters 8 32
Lag of pitch filter 7 28
Total 29 159
8/3/2019 ch5.3 (Vocoders)
23/23
23
VSELP Decoder :
1
0
Long-termFilter state
Codebook1
Codebook2
Pitch
synthesisfilter
Spectralpost filter
Spectralenvelop
(LP)synthesis
filter
Synthetic
Speech
2