MS12-6-Barry3newF

bottlelewdMobile - Wireless

Dec 12, 2013 (3 years and 6 months ago)

88 views

1

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Mobile Systems


Lecture 6


Mobile telephony & speech
coding


COMP28512

Steve Furber & Barry Cheetham

2

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Cellular concept



Frequencies different in adjacent cells



‘Seamless hand
-
over’ as user moves from cell to cell.


To add more users, make cells smaller & reduce power.


Reducing power makes bit
-
errors more likely


need FEC

f3

f1

f2

f3

f2

f2

f3

f1

f1

f2

f1

f3

f1

f1



‘Spatial multiplexing’



Divide city into small areas called
cells



From 0.1 to 35 km in diameter.



Hexagonal shape is hypothetical.



Each cell given a frequency band


e.g. f1, f2, f3



Bands re
-
used when cells are far away.



Users must not transmit ‘too loud’

3

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Base
-
stations & antennas

4

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Tower antenna (with UMTS)

5

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Generations of mobile telecoms standards


0G Radio telephones (non
-
cellular)


1G (1983) Cellular analogue for voice


e.g. AMPS


2G (1991) Cellular digital for voice & slow data; e.g. GSM, IS95


2.5G(

1998) Introduce GPRS (56
-
114 kb/s)


2.75G(

2003
) Add EDGE(E
-
GPRS) (up to 384 kb/s)


3G (

2001
) IMT2000 for speech & faster data
-

UMTS etc


3.5G(

2007
) HSPDA (1.8
-
7.2 Mb/s downlink); UL: 384 kb/s


3.75G (

2010) HSPA+ (DL: 56, UL: 22 Mb/s) etc.


3.95G (?) 3GPP
-
LTE, mobile WIMAX, etc.


4G (?) ITU
-
‘IMT Advanced’

6

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Some Acronyms


AMPS


Advanced (analogue) mobile phone system


GSM


(European) Global system for mobile comms


IS95


USA equivalent of GSM


GPRS

General packet radio system for 2G (56
-
114 kb/s)


EDGE


Enhanced GPRS (


384 kb/s)


IMT2000


International mobile telecomms (3G standard)


UMTS


Universal mobile telecoms system


HSDPA


High speed downlink packet access


HSPA+


High speed packet access


LTE


Long term evolution (from 3G to 4G)


WiMAX
-

Worldwide Interop for Microwave Access


ITU


International telecomms Union


3GPP


3G Partnership Project (ex GSM)


3GPP2
-
3G Partnership Proj 2 (ex IS
-
95 & CDMA2000 in USA)

7

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Multiplexing
-
sharing radio spectrum


FDMA
-
Frequency division multiplexed access
(0G & 1G)


Each transmitter given a different ‘carrier’ frequency


TDMA
-
Time division multiplexed access
(2G
-
GSM)


Each transmitter given a regular time
-
slot


CDMA
-
Code division multiplexed access
(2G
-

IS95 & 3G)


Each transmitter uses same band with a unique code


OFDMA
-
Orthogonal frequency division multiplex access
(4G)


(with MIMO
-
multi input/multi output antennas)


Each transmitter uses several ‘carrier’ frequencies at once.


Packetised transmission compatible with IP


ALL use spatial multiplexing (cells) as well
(apart from 0G)


8

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

GSM
-
TDMA (in 900 MHz band)

My speech/data (114 bits)

Your speech/data (114 bits)

4.615ms

1/4.615m

217
TDM frames/s

Bit
-
rate


114 x 217


24.7 kb/s.

9

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

GSM
-
TDMA observations


Two 25 MHz channels used for 124


8 = 992 users.


FDMA divides 25 MHz into 124 bands (0.2 kHz each).


TDMA divides each band into 8 time
-
slots.


Different slots for base to mob & mob to base (why?)



24.7 kb/s in 114 bit ‘packets’ (217 ‘packets’/s).


Short packets, regular slots.


No ‘contention mode’ & little delay.



Supports
13 kb/s

coded speech or data with FEC.



Some capacity used for synchronisation, signalling etc.

10

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

CDMA


Each bit from coded speech signal is ‘spread’, e.g. to produce



1 = 1 0 1 0 1 1 0 0 0 1 0 1 0 1 0 1 0 1 1 1 0 1 0 1 0 1 0 1 0 1


0 = 0 1 0 1 0 0 1 1 1 0 1 0 1 0 1 0 1 0 0 0 1 0 1 0 1 0 1 0 1 0



Each ‘bit’ becomes a pseudo
-
random sequence of ‘chips’


Transmitted at high chip
-
rate


needs wide bandwidth.



Receiver can recover orig signal if pseudo
-
random sequence is known.


Otherwise transmission will be heard as noise.


All users transmit at the same time in the same frequency band.


But they all use a different sequence


Receivers can recover each bit by a cross
-
correlation process.




Has ‘soft’ capacity limit.


CDMA used by 2G
-
IS95 in USA and 3G everywhere.

11

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Example: CDMA coding


For ‘chips’ 1 & 0, user A transmits 1 volt &
-
1 volt respectively


For ‘1’ send 1
-
1 1
-
1 1 1
-
1
-
1
-
1 1 (code with 10 chips)


For ‘0’ send
-
1 1
-
1 1
-
1
-
1 1 1 1
-
1


Multiply what we receive by 1
-
1 1
-
1 1 1
-
1
-
1
-
1 1 & sum.


For ‘1’, we get +10, for ‘0’, we get
-
10.



User B uses a different code:


‘1’


1 1 1 1 1
-
1
-
1 1
-
1
-
1 ‘0’


-
1
-
1
-
1
-
1
-
1 1 1
-
1 1 1


Multiplying by user B’s code & summing gives +10 or
-
10.


Multiplying by user A’s code & summing gives 0 (both cases)



If A & B transmit together, say ‘1’ & ‘0’, the voltages add.


We receive 0
-
2 0
-
2 0 2 0
-
2 0 2


Mult by A’s code & add gives +10 & mult by B’s code & add gives
-
10.

12

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

OFDM & MIMO


OFDM uses many sinusoidal carriers simultaneously.


Data spread out among them so that if some are not
received, data can be obtained from the others.


MIMO can double the capacity of a radio channel by
having 2 transmit and 2 receive antennas.

13

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Mobile handset


Must provide:


radio functions


access the network


transmit / receive user data


manage radio resources & connections


human interface


microphone, loudspeaker, display & keyboard

14

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)


Designing a mobile phone

15

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

A typical implementation

16

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Typical implementation


Single Chip Digital Baseband Subsystem


TMS320C54x
-

specialised wireless DSP


ARM7TDMI
-

microprocessor


processor cores can be programmed to
support any digital wireless protocol


platform can be used to design systems for
use in any region of the world

17

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Speech transmission


Obtain encoded segments of digitised voice by


Capturing analog voice, low
-
pass filtering, sampling & quantising.


Packetising into typically 20ms (160 sample) blocks


Applying LPC compression to reduce bit
-
rate to 260 bits/block


(Equivalent to 13 kb/s).


Applying FEC in case bit
-
errors occur.


(Increases bit
-
rate to approx 24.7 kb/s)


Interleaving in case bit
-
errors are in ‘bursts’


Adding info. & encrypting for security


Each time the assigned time
-
slot comes around,


Take 114 bits from the encoded segment.


Modulate them onto a sinusoidal carrier of the assigned frequency



Transmit them by applying the resulting voltage to an antenna.


Continue these concurrent processes.

18

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Buffering (2G
-
GSM)

Segments of
compressed
speech

(13 kb/s)


260 bits

every 20 ms



490 bits

every 20 ms

114 bits
every
4.615 ms

Buffer

Apply FEC
etc.

(


㈴⸷2止⽳k

T牡湳浩m

(


㈴⸷.止⽳/

19

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Speech receiver


In each assigned time
-
slot, receive assigned modulated carrier.


Demodulate to extract 114 bits


Accumulate to form coded voice segment.


When a segment is complete,


remove encryption & interleaving.


Apply FEC error
-
correction to obtain 260 bits.


Apply LPC decoder to obtain 160 samples of speech.


Send samples to A to D converter, lowpass filter, amplifier & speaker.



Continue these two concurrent processes.

20

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Uniform quantisation


Each sample of speech x(t) is represented by a binary number x[n].


Each binary number represents a quantisation level.


With
uniform

quantisation there is constant voltage difference


between levels
.

000

111

x(t)

n



110

101

100

011

010

001

1

2

3

4

5

6

7

8

x[n]

7


6


5


4


3





V潬瑳




T

21

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Quantisation error



If samples are rounded, uniform quantisation produces

2
/


]
[


2
/

where
]
[
)
(


]
[







n
e
n
e
nT
x
n
x

unless overflow occurs when magnitude of e[n] may >>

/2.


Overflow is best avoided.


e[n] is quantisation error.

22

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)


Samples e[n] are ‘random’ within

/2.


If x[n] is converted back to analogue form, these samples are
heard as a ‘white noise’ sound added to
x(t).


Noise is an unwanted signal.


White noise is spread evenly across all frequencies.


Sounds like a waterfall or the sea.


Not

a car or house alarm, or a car revving its engine.


Samples e[n] have uniform probability between

/2.


It may be shown that the mean square value of e[n] is:

12
2



Becomes the power of analogue quantisation noise.



Power in Watts if applied to 1 Ohm speaker. Loudness!!

Noise due to uniform quantisation error

23

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Signal
-
to
-
quantisation noise ratio (SQNR)







Measure how seriously signal is degraded by quantisation noise.

(dB.)

decibels
in

power

noise
on
quantisati
power

signal
log
10

SQNR

10











With uniform quantisation, quantisation
-
noise power is

2
/12



Independent of signal power.



Therefore,
SQNR will depend on signal power.



If we amplify signal as much as possible without overflow, for


sinusoidal waveforms with m
-
bit uniform quantiser:


Approximately true for speech also.

SQNR


6m + 1.8 dB.


24

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)


For telephone users with loud voices & quiet voices,


quantisation
-
noise will have same power,

2
/12.





may be too large for quiet voices
, OK for slightly louder ones,


& too small (risking overflow) for much louder voices.


Useful to know over what range of loudness will speech quality


be acceptable to users.



Variation of input levels

000

111

001

volts

OK



瑯漠扩朠景b
煵楥琠癯tce



瑯漠獭慬氠景l
汯畤⁶楣e





25

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Dynamic Range (
Dy
)


.






.
)

(




log

10
10
dB
SQNR
acceptable
gives
which
power
Min
overflow
no
power
signal
possible
Max
Dy










Dy

=

max

possible

SQNR

(dB)



min

acceptable

SQNR

(dB)

















12
/

log
10
12
/


log
10
:
on
quantisati

uniform
For
2
10
2
10
power
Min
power
signal
Max
Dy
This final expression for
Dy

is well worth remembering,


but it only works for uniform quantisation!

26

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Companding for ‘narrow
-
band’ speech




‘Narrow
-
band’ speech is what we hear over telephones.


Normally band
-
limited from 300 Hz to about 3500 Hz.


May be sampled at 8 kHz.



8
-
bits per sample not sufficient for good ‘narrow
-
band’ speech
encoding with uniform quantisation.


Problem lies with setting a suitable quantisation step
-
size

.


One solution is to use instantaneous companding.


Step
-
size adjusted according to amplitude of sample.


For larger amplitudes, larger step
-
sizes used as illustrated next.


‘Instantaneous’ because step
-
size changes from sample to sample.


27

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Non
-
uniform quantisation used for companding

x(t)



t

0001



-
001

0111

-
111

0110

-
110

0101

-
101

x[n]

0100

28

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Implementation of companding


Digitise
x(t)

accurately with uniform quantisation to give x[n]
.


Apply compressor formula to x[n] to give y[n].


Uniformly quantise y[n] using fewer bits


Store or transmit the compressed result.


Passing it thro’
expander
reverses effect of compressor.


As y[n] was quantised, we don’t get x[n] exactly.

Uniform

quantise

(many
bits)



Compressor


Expander

x(t)

x[n]

Transmit

or store

y[n]

x’[n]

Uniform
quantise
(fewer
bits)

29

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Effect of compressor


Increase smaller amplitudes of x[n] & reduce larger ones.


When uniform quantiser is applied, fixed


appears:



smaller in proportion to smaller amplitudes of x[n],


larger in proportion to larger amplitudes.


Effect is non
-
uniform quantisation as illustrated before.


Famous compressor formulas: A
-
law & Mu
-
law.



When there are few bits, like 8, the expander is often
implemented by a ‘look
-
up’ table.


30

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Differential coding




Encode differences between samples
.




Where differences transmitted by PCM this is ‘differential PCM’.



31

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Simplified DPCM coder & decoder

Delay by



1 sample

Quantiser



s[n]



e[n]



Transmit or store samples





s[n

-

1]



e[n] = e[n] + q[n]





Coder

Decoder

Delay by



1 sample





s[n

-

1]



Receive from

channel
or store

e[n]

]
[
ˆ
n
s
32

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Why is speech is well suited to differential coding?




Speech energy biased towards low frequency end of spectrum.




Especially true for voiced, (not so much for unvoiced)




Adjacent speech samples often quite close




As amplitudes of e[n]
<

amplitudes of s[n]


e[n] may be encoded using fewer bits/sample.


33

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Modify simple DPCM diagram in 2 ways



Receiver
: introduce


with

=0.99.


Transmitter
: derive as at the receiver.


^


Instead of s[n]
-

s[n
-
1] , transmit s[n]
-



s
[n
-
1] .

]
[
ˆ
n
s
34

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Decoder drawn backwards

Decoder

Delay by

1 sample







s[n

-

1]



Receive from

channel
or store

e[n]

]
[
ˆ
n
s
Exactly

same
decoder

Delay by

1 sample





s[n
-
1]



e[n]

]
[
ˆ
n
s
35

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Modified DPCM encoder

--





z
-
1



z
-
1

Quantise





Copy of receiver



e[n]



e[n]



s[n]



s[n]



^

^

e[n]

^

+

36

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Practical DPCM encoder

Encoder diagram simplifies to:



z
-
1

Quantise

e[n]

s[n]



s[n]







e[n]



^

^





+

e[n]

^

37

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

32 kb/s ADPCM




DPCM with adaptive quantiser


(adapts its step
-
size according to signal e[n] )




ITU standard for speech coding (G726)


(also for 40, 24 & 16 kb/s)



38

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Linear prediction coding (LPC)




Concept of differential coding described in terms of prediction.



Consider again a differential coder as shown below:

s[n]



s[n
-
1]



Quantiser



e[n]


s[n]

z



-

1



Prediction

Prediction error



e[n]

39

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)



‘Predict’ that
s[n]

will be identical to

DPCM viewed as prediction

]
1
[
ˆ

n
s
]
[

to
prediction

a

be

to
]
1
[
ˆ
Consider
n
s
n
s




Prediction error will be
e[n]
.



Receiver makes same prediction


that next sample
will be identical to current reconstructed speech sample.



Only need to transmit the ‘prediction error’ to allow
receiver to make the necessary correction.



If prediction good,
e[n]

small & fewer bits needed for it.



Use several previous samples for a better prediction:


40

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Better prediction


+



+

+



+



s[n]



s[n]

b
1

b
2



b
3

b
M



z
-
1

z
-
1

z
-
1

z
-
1



e[n
]



Prediction





41

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)



Coeffs b
1
, b
2
, ..., b
M

must be adaptive,



Need a new set of coeffs for each block of 10 or 20 ms.



Calculated by a well known ‘LPC’ algorithm.



Code and transmit e[n] & coeffs b
1
, b
2
, ..., b
M

for each block.


Typically M=10.

Short term linear prediction (LP)

]
[
ˆ
...
]
2
[
ˆ
]
1
[
ˆ


Prediction
2
1
M
n
x
b
n
x
b
n
x
b
M







42

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)



Reconstruct speech s[n] as follows:



Prediction filter now puts back what was subtracted at transmitter

LPC Decoder

+

+

+



s[n]

b
1



b
2



b
3





z
-
1

z
-
1

z
-
1

z
-
1



e[n]

Prediction



+

b
M



43

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)



M between 6 &12.



Error {e[n]} has special properties which allow it to be coded


very efficiently.



Study human speech production mechanism.

Comments on LPC coder/encoder

44

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)



Air forced thro’ vocal cords causes them to vibrate.



Periodic build
-
up & release of pressure produces sound.



"Vocal tract excitation".



If high
-
pass filtered, would appear as a series of pulses:

Mouth

Nose



Air from

lungs



Vocal

cords



Vocal tract

Velum





Excitation (high
-
pass filtered)

T



Time



Volts





Speech
-
like

waveform



Time



Voiced speech (vowels)

45

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)


Voiced speech (vowels):


Tube produces resonances which change as person speaks.


Resonances (formants) determine the vowel sound.




Unvoiced speech (consonants):


Vocal cords do not vibrate.


Turbulent air flow produces “hissing” sound.


Vocal tract excitation is random noise
-
like signal.

Voiced & Unvoiced speech

46

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Effect of prediction at the transmitter




With correctly adapted coeffs, subtracting the prediction at the


transmitter removes resonances (formants).




Remaining ‘prediction error’ (or ‘residual’) signal {e[n]}


becomes high
-
pass filtered excitation signal:




periodic series of pulses (voiced),


or



spectrally white random signal (unvoiced).

47

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Voiced/unvoiced speech & residual (e[n])

e[n]

n

P

P

P

s[n]

P

P

P

n

s[n]

e[n]

n

n

Unvoiced

Voiced

48

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)



DSP procedure: “LP(C)” analysis.



Takes block of say 160 samples of speech (

20 ms).



Coeffs calculated such that energy of {e[n]} minimised.



‘Durbin’s Algorithm’ normally used
.


Deriving prediction coeffs b
1
, b
2
, …, b
M

49

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)


‘LPC
-
10’




2400 b/s coder once widely used in military comms.



Encodes 20 ms speech frames by deriving:



b
1

, b
2

, ..., b

M

with M=10 by LPC analysis,



{e[n]} by calculating & subtracting prediction



Instead of transmitting {e[n]} directly, we send:

(i) Unvoiced/voiced decision (1 bit)

(ii) Amplitude ( a single number: 8 bits say)

(iii) A pitch
-
period (a single number: 8 bits say)

50

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

LPC
-
10 receiver

LPC ‘synthesis’

Coefficients

Speech

{e[n]}

Noise
generator

Impulse

sequence

generator

Pitch
-
period P

V/UV decision

Amplitude

51

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

LPC10 & CELP



LPC
-
10 coder at 2400 b/s is widely used in military
comms.




Another form of LPC, CELP used in mobile
telephony (13 kb/s).



Codebook approach to coding e[n]. Better quality
than LPC10.

52

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Demo in MATLAB


A simple LPC
-
10 coder has been written and tested in MATLAB.


Derives LPC coeffs & prediction error for 20ms segments of speech.


To make things easy, forget about unvoiced speech.


Also use a fixed frequency excitation signal consisting of a series of
pulses at intervals of say 80 samples.


Only the LP filter coeffs and the energy of {e[n]} per 160 sample frame
need be transmitted or stored.


Forget about quantising filter coeffs.

53

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)


Waveform coding and parametric coding.



Waveform coding techniques such as PCM & DPCM try to
preserve exact shape of speech waveform as far as possible.


Simple to understand & implement, but cannot achieve very low
bit
-
rates.


Parametric techniques (e.g. LPC) do not aim to preserve exact
wave
-
shape, & represent features expected to be perceptually
significant by sets of parameters,


i.e. b
i

coefficients & parameters of stylised error signal.



Parametric coding more complicated to understand &
implement than waveform coding, but achieves lower bit
-
rates.

54

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

‘Comfort noise’


Option for not transmitting ‘silence’.


Saves transmission power.


But receiver’s phone may sound ‘dead’


No background noise heard.


So we insert some artificial background noise


‘Comfort noise’.


Not always used, I think.

55

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

‘Low delay’ LPC


A form of LPC which adapts b
1
, b
2

… continuously by
making a small change for each speech sample.


Uses a form of the 'LMS algorithm’.


Changes to b
1
, b
2

… cause e[n] to reduce a little in power.


Encoder must have a ‘built in’ copy of the decoder.


Try to do this for a 2
nd

order predictor


e[n] = s[n]


b
1

s[n
-
1]


b
2

s[n
-
2]


Level of e[n] likely to be reduced if b
1

& b
2

adapted as
follows for each sample:


b
1

=
λ
.b
1



μ
.e[n].s[n
-
1]


b
2

=
λ
.b
2



μ
.e[n].s[n
-
2]


Look up ‘LMS_Algorithm or wait for future lecture.

56

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

‘Low delay LPC’ (cont)



λ (lambda)


0.99: causes b
1

& b
2

to gradually forget any
differences at start
-
up.


Choosing


is tricky. Start small (say 10
-
10
) then increase.


No need to transmit the LPC coeffs separately.


Same adaptations to b
1

and b
2

performed at decoder.


Adaptations are very small for each sample.


Gradually accumulate to make prediction better & better.


Start with b
1

= 1 & b
2

= 0.


LPC coder is initially identical to ADPCM encoder


It should then gradually improve if the adaptation works.


57

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

‘Low delay LPC’ (concl.)


Low delay LPC with much larger values of M can achieve
'good' quality speech at 16 kb/s.


G728


Low delay CELP


Task 2.A.4 is to improve the quality achieved by the 32
kb/s ADPCM encoder developed in Task 2.A.3.


One option is to use low delay LPC with M=2.

58

13 Mar 2012 Lecture 6

COMP28512 (COMP20252)

Summary


Cellular concept & spatial multiplexing.


Generations of mobile telecoms standards.


Multiplexing with FDMA, TDMA, CDMA, OFDMA


Mobile handsets generally employ a general
-
purpose processor
and a DSP


Uniform quantisation noise power:

2
/12



SQNR = 6m + 1.8 dB (M bits)


Dynamic range (in dB)


Speech bit
-
rate compression by non
-
uniform companding.


Differential coding & ADPCM.


Concept of linear prediction.


LPC voice coding using compression technique closely based
on the characteristics of the human voice.


LPC10 illustrated.


CELP (used in mobile telephony at 13 kb/s) mentioned.


Difference between waveform & parametric speech coding.