Error recovery method and apparatus for ADPCM encoded speech

Details Error recovery method and apparatus for ADPCM encoded speech Error recovery method and apparatus for ADPCM encoded speech

1999-01-20

2003-06-10

Decady, Albert (Department: 2133)

Error detection/correction and fault detection/recovery

Pulse or data error handling

Error count or rate

C375S244000, C375S351000

Reexamination Certificate

active

06578162

ABSTRACT:

I. BACKGROUND OF THE INVENTION
The present invention relates generally to error recovery for encoded speech in a digital communication system, and more specifically, to error recovery for speech signals encoded using adaptive differential pulse code modulation (ADPCM).
Encoders and decoders are commonly employed in communication systems for the purpose of compressing and decompressing speech signals. Adaptive Differential Pulse Code Modulation (ADPCM) describes a form of encoding speech signals in a digital communication system in which compression ratios of 2:1 or even 4:1, with respect to 8-bit compressed PCM samples, can be achieved with relatively low levels of complexity, delay, and speech degradation. In the last few years, this form of encoding has been incorporated into various Personal Communication System (PCS) standards, including the Japanese Personal Handi-Phone System (PHS) and European Digital European Cordless Telecommunications (DECT) standards. It has also become the de facto standard in the United States for the coding of speech in cordless telecommunications systems. The particular form of ADPCM employed in these systems is described in CCITT Recommendation G.726, “40, 32, 24, 16 kbit/s ADAPTIVE DIFFERENTIAL PULSE CODE MODULATION (ADPCM),” Geneva, 1990 (hereinafter referred to as “CCITT Recommendation G.726”), which is hereby fully incorporated by reference herein as though set forth in full.
A problem arises because this G.726 standard was developed for terrestrial wireline applications, not radio frequency (RF) systems employing dispersive channels, such as the foregoing PHS and DECT cordless systems, and wireless systems, such as digital PCS, in which the channel error rate experienced is typically much greater due to factors such as interference from other users and multipath fading. More specifically, a G.726 ADPCM decoding and encoding system quickly degrades when subjected to such error rates. Consequently, audible “clicks” or “pops” occur when speech passing through such a system is played over a speaker. This problem stems from the structure of the G.726 ADPCM encoder and decoder, which will now be explained.
A block diagram of a G.726 compliant encoder is illustrated in FIG.
1
. As can be seen, this encoder comprises Input PCM Format Conversion Block
1
, Difference Signal Computation Block
2
, Adaptive Quantizer
3
, Inverse Adaptive Quantizer
4
, Reconstructed Signal Calculator
5
, Adaptive Predictor
6
, Tone And Transition Detector
7
, Adaptation Speed Control Block
8
, and Quantizer Scale Factor Adaptation Block
9
, coupled together as shown. This figure and the following explanation is taken largely from CCITT Recommendation G.726. This encoder receives as input pulse-code modulated (PCM) speech samples, s(k), and provides as output ADPCM samples I(k). In one implementation, in which the mode of transmission is analog transmission, the PCM samples, s(k), are uniform PCM samples. In one example of this implementation, the PCM samples are 14-bit uniform samples which range from −8192 to +8191. In this implementation, Block
1
can be eliminated since the PCM samples are already in a uniform format. In another implementation, in which the mode of transmission is digital transmission, the PCM samples are A-law or &mgr;-law samples. In one example of this implementation, the PCM samples are compressed 8-bit samples. The output ADPCM samples, I(k), are generated from an adaptively quantized version of the difference signal, d(k), which is the difference between the uniform PCM signal, s
1
(k), and an estimated signal, s
e
(k), provided by Block
6
. In these variables, k is the sampling index. In one embodiment, the sampling interval is 125 &mgr;s. A basic assumption is that s
e
(k) can be precisely recreated at the decoder in order to regenerate the speech signal from received values of I(k).
Optional block
1
converts the input signal s(k) from A-law or &mgr;-law format to a uniform PCM signal s
1
(k). Block
2
outputs a difference signal, d(k), equal to s
1
(k)−s
e
(k). Block
3
is a non-uniform adaptive quantizer used to quantize d(k) using an adaptively quantized scale factor, y(k), output from Block
9
. This quantizer operates as follows. First, the input d(k) is normalized using the following equation: log
2
|d(k)|−y(k). Then, a value for the output I(k)is determined responsive to this normalized input. In one embodiment, in which the output is selected to be at the rate 32 kbit/s, each output value is four bits, three bits for the magnitude and one bit for the sign, specifying one of sixteen quantization levels as determined by the following table:
Normalized quantizer input
Normalized quantizer output
range: log
2
|d(k) − y(k)|
|I(k)|
log
2
|d
q
(k)| − y(k)
[4.31, +∞]
15
4.42
[4.12, 4.31)
14
4.21
[3.91, 4.12)
13
4.02
[3.70, 3.91)
12
3.81
[3.47, 3.70)
11
3.59
[3.22, 3.47)
10
3.35
[2.95, 3.22)
9
3.09
[2.64, 2.95)
8
2.80
[2.32, 2.64)
7
2.48
[1.95, 2.32)
6
2.14
[1.54, 1.95)
5
1.75
[1.08, 1.54)
4
1.32
[0.52, 1.08)
3
0.81
[−0.13, 0.52)
2
0.22
[−0.96, −0.13)
1
−0.52
(−∞, −0.96)
0
−∞
Block
4
provides a quantized version of the difference signal, d
q
(k), from I(k) in accordance with the foregoing table. More specifically, through an inverse quantization process, a normalized quantizer output in the rightmost column of the table is selected based on the value of I(k). Then, referring to this value as N.O., d
q
(k) is determined using the following equation: |d
q
(k)|=2
|N.O.|+y(k)
, in which N.O. is the normalized quantizer output. Because of quantization error, the signal d
q
(k) will typically differ from d(k).
Block
9
adaptively computes the scale factor, y(k), in part based on past values of y(k). More specifically, a fast (unlocked) scale factor y
u
(k) is computed using the following equation: y
u
(k)=(1−2
−5
)y(k)+2
−5
W[I(k)]. For 32 kbit/s ADPCM, the function W[I(k)] is defined as follows:
|I(k)|
7
6
5
4
3
2
1
0
W[I(k)]
70.13
22.19
12.38
7.00
4.00
2.56
1.13
−0.75
Thus, higher magnitude values of I(k) are weighted significantly more heavily than lower magnitude values of I(k).
A slow (locked) scale factor y
l
(k) is derived from y
u
(k) using the following equation: y
l
(k)=(1−2
−6
)y
l
(k−1)+2
−6
y
u
(k). The fast and slow scale factors are then combined to form y(k) using the adaptive speed control factor a
1
(k) provided from Block
8
, where 0≦a
1
(k)≦1. The following equation describes the specific relationship between these variables: y(k)=a
1
(k)y
u
(k−1)+[1−a
1
(k)]y
l
(k−1).
The parameter a
1
(k) provided by Block
8
can assume values in the range [0,1]. It tends towards unity for speech signals, and towards zero for voiceband data signals. To compute this parameter, two measures of the average magnitude of I(k), d
ml
(k) and d
ms
(k), are computed using the following equations:
d
ms
(
k
)=(1−2
−5
)
d
ms
(
k
−1)+2
−5
F[I
(
k
)]
d
ml
(
k
)=(1−2
−7
)
d
ml
(
k
−1)+2
−7
F[I
(
k
)]
For 32 kbit/s ADPCM, F[I(k)] is defined by:
|I(k)|
7
6
5
4
3
2
1
0
F[I(k)]
7
3
1
1
1
0
0
0
Thus, d
ms
(k) is a relatively short-term average of F[I(k)], and d
ml
(k) is a relatively long-term average of F[I(k)]. Using these two averages, the variable a
p
(k) is computed. The variable a
p
(k) tends towards the value of 2 if the difference between d
ms
(k) and d
ml
(k) is large (average magnitude of I(k) changing) and tends towards the value of 0 if the difference is small (average magnitude of I(k) relatively constant). Further details about the computation of

Affiliated with

Also associated with

Rating

Error recovery method and apparatus for ADPCM encoded speech does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Error recovery method and apparatus for ADPCM encoded speech, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Error recovery method and apparatus for ADPCM encoded speech will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3090318