CA1222570A - Predictive code conversion method capable of forcibly putting a feedback loop in an inactive state - Google Patents
Predictive code conversion method capable of forcibly putting a feedback loop in an inactive stateInfo
- Publication number
- CA1222570A CA1222570A CA000459932A CA459932A CA1222570A CA 1222570 A CA1222570 A CA 1222570A CA 000459932 A CA000459932 A CA 000459932A CA 459932 A CA459932 A CA 459932A CA 1222570 A CA1222570 A CA 1222570A
- Authority
- CA
- Canada
- Prior art keywords
- encoder
- signal
- decoder
- succession
- predictor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired
Links
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M3/00—Conversion of analogue values to or from differential modulation
- H03M3/04—Differential modulation with several bits, e.g. differential pulse code modulation [DPCM]
- H03M3/042—Differential modulation with several bits, e.g. differential pulse code modulation [DPCM] with adaptable step size, e.g. adaptive differential pulse code modulation [ADPCM]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Transmission Systems Not Characterized By The Medium Used For Transmission (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
Abstract of the Disclosure:
In a predictive encoder for encoding an encoder input signal into an encoder output signal and having in a feedback loop an adaptive predictive filter which comprises a plurality of taps assigned with adaptive prediction coefficients and produces a predictive signal predictive of the encoder input signal, an encoder control circuit monitors products between the coefficients and tap signals derived from the taps so as to render each product forcibly zero when an absolute value of each product is not greater than a predetermined value.
The feedback loop is transiently put into an inactive state when the predictive signal becomes zero as a result of all zeros of the products. A predictive decoder is also operated by the use of a decoder control circuit similar to the encoder control circuit and cooperating with an adaptive predictive filter of the decoder.
Each adaptive predictive filter may be either of an FIR type or of an IIR type.
In a predictive encoder for encoding an encoder input signal into an encoder output signal and having in a feedback loop an adaptive predictive filter which comprises a plurality of taps assigned with adaptive prediction coefficients and produces a predictive signal predictive of the encoder input signal, an encoder control circuit monitors products between the coefficients and tap signals derived from the taps so as to render each product forcibly zero when an absolute value of each product is not greater than a predetermined value.
The feedback loop is transiently put into an inactive state when the predictive signal becomes zero as a result of all zeros of the products. A predictive decoder is also operated by the use of a decoder control circuit similar to the encoder control circuit and cooperating with an adaptive predictive filter of the decoder.
Each adaptive predictive filter may be either of an FIR type or of an IIR type.
Description
~22~1s70 Bac~ground of the Invention This invention relates to a method of carrying out conversion between first and second signal successions in a differential pulse code modulation (DPCM) system and to an encoder and a decoder for use in carrying out the conversion.
More particularly, this invention is applicable to an adaptive differential pulse code modulation (ADPCM) system. It is to be noted here that the conversion method may be an encoding method and/or a decoding method and that the encoder encodes the first signal succession into the second signal succession while the decoder decodes the second signal succession into the first signal succession.
A conventional conversion method is described by Takao Nishitani et al in ICASSP 82 Proceedings, Volume 2 (May 1982), pages ~:~
960-963, under the title of "A 32 kb/s Toll Quality ADPCM Codec using a Single Chlp Signal Processor."
In the conventional conversion method, adaptive prediction is carried out in each of an encoder and a decoder by the use of an adaptive predictor so as to successively predict a current or predictive signal from preceding or past signals. The encoder encodes a first signal succession with reference to the predictive signals into a second signal succession which is produced in the form of predictive error signals between the first signal succession and the predictive signals. On the other hand, the decoder reproduces the first signal succession in response to the received second signal succession with reference to the predictive signals derived from the received second signal succession.
Anyway, each predictive signal is adaptively varied in accordance with an input signal supplied to each predictor, so as to enable precise prediction.
As known in the art, the conversion method is very effective because electric power of the second electric signal succession can be reduced as compared with that of the first signal succession, when the first signal succession carries a signal, such as an audio signal, having a strong correlation.
It should be noted here that operation of the encoder is subject ko an influence of an initial state for a long time after the beginning of operation and such an influence remains unextinct. This is because .,~ .
..
the encoder has a feedback loop including the predictor.
As a result, encoding precision or quali-ty of the second signal succession is often degraded in dependence on the initial state. For example, when the initial state of the encoder is different at the beginning of operation from that at the beginning of-the previous operation, encoding quality is also varied from one another.
Such variation of the encoding quality adversely affects decoding operation of the decoder.
Summary of the Invention:
. . :
It is an object of this invention to provide a method of carrying out conversion between two signal successions by the use of an adaptive predictor, wherein an influence of an initial state can instantly be removed.
It is another object of this invention to provide a method of the type described, which can avoid degradation of encoding quality.
It is a further object of this invention to provide a predictive encoder which is for use in the method and which can instantly remove an influence result-ing from variation of initial states thereof.
It is another object of this invention to provide a predictive decoder which is communicable with the above-mentioned encoder.
According to this invention, a method is for carrying out conversion between a succession of first signals and a succession of second signals by the use of an adaptive pFedictor which is responsive to the 7~
second signal succession to produce a succession of predictive signals predictive of the firs-t signals, respectively. The adaptive predictor comprises tap signal producing means having a plurality of taps assigned with prediction coefficients, respectively, for producing a plurality of tap signals in response to the second signal succession, respectively, coefficient signal producing means for adaptively producing a plurality of coefficient signals representative of the prediction coefficien-ts, respectively, a plurality of multipliers responsive to the tap signals and the coefficient signals for calculating products therebetween to produce product signals representative of the products, respectively, and output means responsive to the product signals for producing the predictive signal succession.
The me-thod comprises the step of moni-toring the products to render each of the products zero when the each product is not greater than a predetermined value.
According to another aspect, the invention provides an encoder for carrying out the inventive method by encoding said first signal succession into said second signal succession, wherein said encoder comprises: encoder monitoring means coupled to said multipliers for monitoring said products -to render each of said products zero when said each product is not greater than said predetermined value.
According to a further aspect, the invention provides a decoder for carrying out the inventive method by decoding said second signal succession into said Eirst signal succession, ~ - 4a -wherein said decoder comprises: decoder monitoring means coupled to said multipliers for monitoring said products to render each of said product zero when said each product is not greater than said predetermined value.
~rief Description of the Drawing:
Figure 1 is a block diagram of a conventional encoding device;
Figure 2 is a block diagram of a conventional decoding device operable in cooperation with the encoding device illust-rated in Figure 1;
Figure 3 is a block diagram of an encoding device according to a preferred embodim~ent of this invention;
Figure 4 is a block diagram of a part of the encoding device illustrated in Figure 3; and ~L2~2S7~
Fig. 5 is a block diagram of a decoding device which is operable in cooperation with the encoding device illustrated in Fig. 3.
Description of the Preferred Embodiments_ Referring to Figs. 1 and 2, a conventional adaptive differential pulse code modulatlon (ADPCM) system will be described at first for a better understanding of this invention. The system comprises an encoding device and a decoding device which are illustrated in Figs.
1 and 2, respectively, and which are exemplified by a predictive encoder 11 and a predictive decoder 12, respectively. The predictive encoder and decoder 11 and 12 may be called ADPCM encoder and decoder, respec-tively. In Fig. 1, the encoder 11 is supplied with a succession of first digital signals x which may be referred to as a successlon of encoder input signals, although only a current one of the first digital signals at a current time instant ~ is represented by Xj in Fig. 1. Similar representation will be used in connection with any other signals, though not explicitly described for each signal. A subtractor 14 subtracts a local output signal Xj of an adaptive predictive filter 16 from the first digital signal x; to produce an error signal ej representative of a difference between the encoder input signal x; and the local output signal x;. The error signal ej may be called a predictive error signal~ as will become clear as the description proceeds. The adaptive predictive filter 16 may be ., ~
~2~5q~
referred to as an encoder predictor.
The predictive error signal ej i-s quantized by a quantizer 18 into a quantized predictive error signal ej in a well-known manner. The quantized predictive error signal ej is delivered as an encoder output signal to an additional encoder 19 which follows the encoder 11. A succession of the encoder output signals may be termed a succession of second digital signals and is subjected to variable length encoding or the like in the additional encoder 19 to be delivered as a differ-ential pulse code modulation (DPCM) signal Uj to a transmission line (not shown).
On the other hand, the quantized predictive error signal êj is also delivered from the quantizer 18 to the adaptive predictive filter 16. In the manner well known in the art, the adaptive predictive filter 16 is provided with a plurality of prediction coefficients which are adaptively modified in accordance with the quantized predictive error signal êj. Moreover, the adaptive predictive filter 16 produces a succession of predictive signals predictive of the encoder input signal succession x with reference to the adaptively modified prediction coefficients. More specifically, when supplied with the quantized predictive error signal êj at the current time instant i~ the adaptive predictive filter 16 produces a current one (Xj) of the predictive signals that is predictive of the current encoder input signal x; and calculates the following predictive signal ~L222S7(3 xj+l at the following time instant (j + 1). At any rate, each predictive signal is sent to-the subtractor 14 as each local output signal.
As shown in Fig. 1, the encoder 11 forms a feedback system comprising the adaptive predictive filter 16 in a feedback loop. The predi-ction coefficients of the adaptive predictive filter 16 are adaptively modified in response to the quantized predictive error signal êj. This implies that the feedback system is not readily put into a stable state. Therefore, when the encoder 11 begins to operate by supply of electric power, the operation is subject to an influence of an initial state at the beginning of the operation. The influence of the initial state lasts for a long time in the feedback loop and remains unextinct, as mentioned in the preamble of the instant specification. This gives rise to a variation of encoding quality in dependence on a difference of the initial state.
In Fig. 2, the conventional decoder 12 is preceded by an additional decoder,22 supplied with a reception signal through a transmission line (not shown). The reception signal is equivalent to the DPCM signal U
(Fig. 1) if no error takes place in the DPCM signal during transmission and is therefore denoted at Uj like the DPCM signal. The reception signal Uj is decoded - by the additional decoder 22 into a reproduced error signal which is a reproduction of the quantized predictive error signal êj (Fig. 1) and which is therefore designated , by the same reference symbol as the quantized predictive error signal ej. The reproduced error signal ej thus corresponds to the quantized predictive error signal ej and may therefore be called the second digital signal like in the quantized predicLive error signal ej.
The reproduced error signal êj is given to the decoder 12 as a decoder input signal and is delivered to an adder 24 and an adaptive predictive filter 26 which may be referred to as a decoder predictor. The decoder predictor 26 produces a reproduced predictive signal Xj at a current time instant ~. The reproduced predictive signal x; is added to the reproduced error signal êj by the adder 24 to be produced as a reproduced signal Xj which is a reproduction of the encoder input signal Xj (Fig. 1). The reproduced signal Xj may be called the first digital signal like the encoder input signal Xj and appears as a decoder output signal.
Responsive to the reproduced error signal êj at the current time instant ~, the decoder predictor 26 adaptively modifies prediction coefficients assigned thereto to calculate the next following reproduced predic-tive signal xj+l for the next subsequent time instant (j ~ 1).
Operation of the decoder 12 is dependent on that of the encoder 11. Accordingly, the decoder 12 is also subjected to the influence of the initial state of the encoder 11. As a result, similar problem arises as regards the decoder 12 also.
,~
257~) It will be assumed that predictive encoding and decoding as mentioned above are often repeated plural times through a plurality of sets, each comprising adaptive predictive encoder and decoder. In this case, initial internal states of each set may be different from those of the other sets. Under the-circumstances, degradation of the encoding quality is accumulated at every set and becomes serious.
Referring to Fig. 3, an encoding device according to a preferred embodiment of thls invention is applicable to the ADPCM system and comprises similar parts and deals with signals, ail of which will be designated by like reference numerals and symbols. The illustrated encoder 11 is similar to that illustrated in Fig. 1 except that an encoder control circuit 32 is coupled to the adaptive predictive filter 16. It should be noted here that the influence of the initial state can be removed by putting the adaptive predictive filter 16, namely, the feedback loop into a transient inactive state. For this purpose, the encoder control circuit 32 monitors the encoder predictor 16 in a manner to be described later.
In Fig. 3, the encoder 11 encodes the encoder input signal x; into the quantized predictive error signal êj supplied as the encoder output signal to the additional encoder 19. The encoder output signal is subjected to another encoding in the additional encoder 19 to be sent as the DPCM signal Uj to a transmission ~ " .
~2~70 line (not shown).
Referring to Fig. 4 afresh together with Fig.
3, the illustrated adaptive predictive filter 16 is of a Finite Impulse Response (FIR) type and comprises a delay circuit 34 comprising first through (N + l)-th taps and first through N-th delay units 361 to 36N each of which is connected between two adjacent ones of the taps, where N is representative of a natural number of, for example, ten. The quantized predictive error signal êj is successively delayed by the first through the N-th delay units 361 to 36N and appears as first through (N + l)-th tap signals Dl to DN+l through the first through the (N + l)-th taps, respectively. There~
fore, the delay circuit 34 may be named a tap signal producing circuit.
The quantized predictive error signal êj is supplied to a gain control circuit 38 for producing first through (N + l)-th coefficient signals Cl to CN+
representative of first through (N + l)-th prediction coefficients, respectively. The first through the (N + l)-th prediction coefficients are assigned to the first through the (N + l)-th taps, respectively, and are adaptively modified under control of the gain control circuit 38 in response to the quantized predictive error signal êj. As a result, the first through the (N + l)-th coeffi-cient signals Cl to CN+l are modified in accordance with the quarltized predictive error signal êj. The first through the (N + l)-th coefficient signals C
7: .
~ ;22;~7~
to C~+1 are in one-to-one correspondence to the first through the ~N + l)-th tap signals Dl to DN+l, respectively.
First through (N + 1)-th multipliers 411 to 41N+l multiply the first through the (N + 1)-th tap signals Dl to DN+l by the first through the (N + l)-th coefficient signals Cl to CN+i to calculate first through (N + l)-th products, respectively. The first through the (N + l)-th products are delivered as first through (N + l)-th product signals to an adder circuit 43, respec-tively. An i-th one of the first through the (N + l)-th product signals Pl to PN+l can be represented by:
Ci X Di.
Thus, each product can be specified by a prescribed number of values between a maximum and a minimum value, -~
both inclusive.
The adder circuit 43 sums up the first through the (N + l)-th products Pl to PN+l and produces the predictive signal Xj given by:
N+l Xj - i~ Ci x Di.
If an audio or speech signal is carried by the encoder input signal succession x, a pause or a quiescent state inevitably occurs in the encoder input signal succession x. Under the circumstances, each of the first through the (N + l)-th products becomes small during the quiescent state but the predictive signal Xj does not easily become zero.
:
~L;22%~
More particularly, let each of the tap signals Di and the coefficient signals Ci have a predetermined length of, for example, four bits. In this event, each product signal Pi can be expressed by eight bits. Each product Pi takes zero when either the tap signal Di or the coefficient signal Ci b-ecomes equal to zero.
In order to render the feedback loop into the inactive state, all of the first through the (N + l)-th product signals Pl to PN+l should be rendered zero. However, the first through the (N + l)-th product signals Pl to PN+l are not readily rendered zero at the same time during the quiescent state of the encoder input signal succession x. Thus, the feedback loop is not readily put into the inactive state.
In Fig. 4, the illustrated control circuit 32 comprises a control portion 44 supplied with the first through the (N + l)-th product signals Pl to PN+l and a threshold circuit 45 for producing a threshold signal TH representative of a threshold value. The control portion 44 monitors the first through the (N + l)-th product signals Pl to PN+l with reference to the threshold signal TH. More specifically, if the value of each product Pi is smaller than the threshold value, the control portion 44 renders the value of each product forcibly zero. Consequently, it is possible to put the adaptive predictive filter 16 into the inactive state by rendering all of the products zero during the quiescent state. For this purpose, the control portion ' , 44 supplies the adder circuit 43 with control signals CTL representative of results of monitor-ing. Production of the control signals CTL can be possible by the use of usual comparators. The control signals CTL are used in the adder circuit 43 to individually gate the first through the (N + l)-th product- signals Pl to PN+l, respec-tively. Therefore, the adder circuit 43 may be a combination of gate circuits and an adder.
As mentioned above, the adaptive predictive filter 16 renders the predictive signal x; zero during the quiescent stateO As a result, the feedback loop is transiently put into the inactive state and the influ-ence of initial state can be removed by shutting off the feedback loop in effect.
Operation of the con rol circuit 32 will be described more in detail. Let each of the coefficient signals Ci and the tap signals Di have a predetermined length of, for example, four bits. In this event, each product signal Pi can be expressed by eight bits and therefore~divisible into a higher significant part of four bits and a lower significant part of four bits.
In addition, let each product be expressed by a fixed point representation and the higher significant part alone be produced from each multiplier as each product signal Pi with the lower significant part truncated.
If Ci and Di are equal to "0011" and "0001,"
the produGt becomes equal to "0000." Therefore, the threshold value may be equal to "0000."
~/
~2~S~70 On the other hand, it will be assumed that each tap signal Di is represented in the notation of two's complement.
If the tap signal Di is equal to minus 0001, the tap signal ~i is represented as "1111" in the two's complement notation. As a result, the product is rendered into "1111" which corresponds to the decimal number -1, not zero. When each product can take a negative value, such as -1, the product signal Pi cannot become zero. As a result, the predictive signal xj also cannot become equal to zero despite the fact that each product is very small.
In this e~ent, the predictive signal xj can become equal to zero by giving the threshold circuit 45 the threshold value of "1111 n and by rendering each product signal Pi forcibly zero when each product is equal to or less than "1111." In this example, each product signal Pi is rendered zero only when each product takes ~1111 . n Thus, an absolute value of each product may be monitored by the control circuit 32~
Likewise, each product signal Pi does not become zero during the quiescent state when each product signal Pi has a too long length. In this event, the control circuit 32 may monitor an absolute value of each product and renders the absolute value zero when the absolute value is equal to or smaller than the threshold value predetermined for each product.
Anyway, the ~eedback loop, namely, the adaptive predic-tive filter 16 can be put into the inactive state ~;,`
257~
during the quiescent state by the use of the control circuit 32.
Referring to Fig. 5, a decoding device is operable in cooperation with the encoding device described in conjunction with Figs. 3 and 4 and is similar to that illustrated in Fig. 2 except t-hat the decoder predictor 26 included in the decoder 12 is coupled to a decoder control circuit 50 which may be called a decoder control circuit. In this connection, the control circuit 32 (Fig. 3) may be named an encoder control circuit. The decoder control circuit 50 is similar in structure and operation to the encoder control circuit 32 illustrated in Fig. 4 and therefore monitors each product signal Pi to render each product forcibly zero when the absolute value of each product is not greater than a threshold value which may be equal to that determined in the encoder control circuit 32.
While this invention has thus far been described in conjunction with a preferred embodiment thereof, it will readily be possible for those skilled in the art to put this invention into practice in various other manners. For example, each of the encoder and the decoder predictors 16 and 26 may be of an Infinite Impulse Response (IIR) type. In this event, each predictor 16 and 26 (Fig. 4) is supplied with a sum of the predictive signal Xj and the quantized predictive error signal ej through an additional adder, instead of the quantized predictive error signal êj. Each of the encoder and the decoder Y'' ~L222~;70 predictors may be a combination of an IIR type and an FIR type.
,,'
More particularly, this invention is applicable to an adaptive differential pulse code modulation (ADPCM) system. It is to be noted here that the conversion method may be an encoding method and/or a decoding method and that the encoder encodes the first signal succession into the second signal succession while the decoder decodes the second signal succession into the first signal succession.
A conventional conversion method is described by Takao Nishitani et al in ICASSP 82 Proceedings, Volume 2 (May 1982), pages ~:~
960-963, under the title of "A 32 kb/s Toll Quality ADPCM Codec using a Single Chlp Signal Processor."
In the conventional conversion method, adaptive prediction is carried out in each of an encoder and a decoder by the use of an adaptive predictor so as to successively predict a current or predictive signal from preceding or past signals. The encoder encodes a first signal succession with reference to the predictive signals into a second signal succession which is produced in the form of predictive error signals between the first signal succession and the predictive signals. On the other hand, the decoder reproduces the first signal succession in response to the received second signal succession with reference to the predictive signals derived from the received second signal succession.
Anyway, each predictive signal is adaptively varied in accordance with an input signal supplied to each predictor, so as to enable precise prediction.
As known in the art, the conversion method is very effective because electric power of the second electric signal succession can be reduced as compared with that of the first signal succession, when the first signal succession carries a signal, such as an audio signal, having a strong correlation.
It should be noted here that operation of the encoder is subject ko an influence of an initial state for a long time after the beginning of operation and such an influence remains unextinct. This is because .,~ .
..
the encoder has a feedback loop including the predictor.
As a result, encoding precision or quali-ty of the second signal succession is often degraded in dependence on the initial state. For example, when the initial state of the encoder is different at the beginning of operation from that at the beginning of-the previous operation, encoding quality is also varied from one another.
Such variation of the encoding quality adversely affects decoding operation of the decoder.
Summary of the Invention:
. . :
It is an object of this invention to provide a method of carrying out conversion between two signal successions by the use of an adaptive predictor, wherein an influence of an initial state can instantly be removed.
It is another object of this invention to provide a method of the type described, which can avoid degradation of encoding quality.
It is a further object of this invention to provide a predictive encoder which is for use in the method and which can instantly remove an influence result-ing from variation of initial states thereof.
It is another object of this invention to provide a predictive decoder which is communicable with the above-mentioned encoder.
According to this invention, a method is for carrying out conversion between a succession of first signals and a succession of second signals by the use of an adaptive pFedictor which is responsive to the 7~
second signal succession to produce a succession of predictive signals predictive of the firs-t signals, respectively. The adaptive predictor comprises tap signal producing means having a plurality of taps assigned with prediction coefficients, respectively, for producing a plurality of tap signals in response to the second signal succession, respectively, coefficient signal producing means for adaptively producing a plurality of coefficient signals representative of the prediction coefficien-ts, respectively, a plurality of multipliers responsive to the tap signals and the coefficient signals for calculating products therebetween to produce product signals representative of the products, respectively, and output means responsive to the product signals for producing the predictive signal succession.
The me-thod comprises the step of moni-toring the products to render each of the products zero when the each product is not greater than a predetermined value.
According to another aspect, the invention provides an encoder for carrying out the inventive method by encoding said first signal succession into said second signal succession, wherein said encoder comprises: encoder monitoring means coupled to said multipliers for monitoring said products -to render each of said products zero when said each product is not greater than said predetermined value.
According to a further aspect, the invention provides a decoder for carrying out the inventive method by decoding said second signal succession into said Eirst signal succession, ~ - 4a -wherein said decoder comprises: decoder monitoring means coupled to said multipliers for monitoring said products to render each of said product zero when said each product is not greater than said predetermined value.
~rief Description of the Drawing:
Figure 1 is a block diagram of a conventional encoding device;
Figure 2 is a block diagram of a conventional decoding device operable in cooperation with the encoding device illust-rated in Figure 1;
Figure 3 is a block diagram of an encoding device according to a preferred embodim~ent of this invention;
Figure 4 is a block diagram of a part of the encoding device illustrated in Figure 3; and ~L2~2S7~
Fig. 5 is a block diagram of a decoding device which is operable in cooperation with the encoding device illustrated in Fig. 3.
Description of the Preferred Embodiments_ Referring to Figs. 1 and 2, a conventional adaptive differential pulse code modulatlon (ADPCM) system will be described at first for a better understanding of this invention. The system comprises an encoding device and a decoding device which are illustrated in Figs.
1 and 2, respectively, and which are exemplified by a predictive encoder 11 and a predictive decoder 12, respectively. The predictive encoder and decoder 11 and 12 may be called ADPCM encoder and decoder, respec-tively. In Fig. 1, the encoder 11 is supplied with a succession of first digital signals x which may be referred to as a successlon of encoder input signals, although only a current one of the first digital signals at a current time instant ~ is represented by Xj in Fig. 1. Similar representation will be used in connection with any other signals, though not explicitly described for each signal. A subtractor 14 subtracts a local output signal Xj of an adaptive predictive filter 16 from the first digital signal x; to produce an error signal ej representative of a difference between the encoder input signal x; and the local output signal x;. The error signal ej may be called a predictive error signal~ as will become clear as the description proceeds. The adaptive predictive filter 16 may be ., ~
~2~5q~
referred to as an encoder predictor.
The predictive error signal ej i-s quantized by a quantizer 18 into a quantized predictive error signal ej in a well-known manner. The quantized predictive error signal ej is delivered as an encoder output signal to an additional encoder 19 which follows the encoder 11. A succession of the encoder output signals may be termed a succession of second digital signals and is subjected to variable length encoding or the like in the additional encoder 19 to be delivered as a differ-ential pulse code modulation (DPCM) signal Uj to a transmission line (not shown).
On the other hand, the quantized predictive error signal êj is also delivered from the quantizer 18 to the adaptive predictive filter 16. In the manner well known in the art, the adaptive predictive filter 16 is provided with a plurality of prediction coefficients which are adaptively modified in accordance with the quantized predictive error signal êj. Moreover, the adaptive predictive filter 16 produces a succession of predictive signals predictive of the encoder input signal succession x with reference to the adaptively modified prediction coefficients. More specifically, when supplied with the quantized predictive error signal êj at the current time instant i~ the adaptive predictive filter 16 produces a current one (Xj) of the predictive signals that is predictive of the current encoder input signal x; and calculates the following predictive signal ~L222S7(3 xj+l at the following time instant (j + 1). At any rate, each predictive signal is sent to-the subtractor 14 as each local output signal.
As shown in Fig. 1, the encoder 11 forms a feedback system comprising the adaptive predictive filter 16 in a feedback loop. The predi-ction coefficients of the adaptive predictive filter 16 are adaptively modified in response to the quantized predictive error signal êj. This implies that the feedback system is not readily put into a stable state. Therefore, when the encoder 11 begins to operate by supply of electric power, the operation is subject to an influence of an initial state at the beginning of the operation. The influence of the initial state lasts for a long time in the feedback loop and remains unextinct, as mentioned in the preamble of the instant specification. This gives rise to a variation of encoding quality in dependence on a difference of the initial state.
In Fig. 2, the conventional decoder 12 is preceded by an additional decoder,22 supplied with a reception signal through a transmission line (not shown). The reception signal is equivalent to the DPCM signal U
(Fig. 1) if no error takes place in the DPCM signal during transmission and is therefore denoted at Uj like the DPCM signal. The reception signal Uj is decoded - by the additional decoder 22 into a reproduced error signal which is a reproduction of the quantized predictive error signal êj (Fig. 1) and which is therefore designated , by the same reference symbol as the quantized predictive error signal ej. The reproduced error signal ej thus corresponds to the quantized predictive error signal ej and may therefore be called the second digital signal like in the quantized predicLive error signal ej.
The reproduced error signal êj is given to the decoder 12 as a decoder input signal and is delivered to an adder 24 and an adaptive predictive filter 26 which may be referred to as a decoder predictor. The decoder predictor 26 produces a reproduced predictive signal Xj at a current time instant ~. The reproduced predictive signal x; is added to the reproduced error signal êj by the adder 24 to be produced as a reproduced signal Xj which is a reproduction of the encoder input signal Xj (Fig. 1). The reproduced signal Xj may be called the first digital signal like the encoder input signal Xj and appears as a decoder output signal.
Responsive to the reproduced error signal êj at the current time instant ~, the decoder predictor 26 adaptively modifies prediction coefficients assigned thereto to calculate the next following reproduced predic-tive signal xj+l for the next subsequent time instant (j ~ 1).
Operation of the decoder 12 is dependent on that of the encoder 11. Accordingly, the decoder 12 is also subjected to the influence of the initial state of the encoder 11. As a result, similar problem arises as regards the decoder 12 also.
,~
257~) It will be assumed that predictive encoding and decoding as mentioned above are often repeated plural times through a plurality of sets, each comprising adaptive predictive encoder and decoder. In this case, initial internal states of each set may be different from those of the other sets. Under the-circumstances, degradation of the encoding quality is accumulated at every set and becomes serious.
Referring to Fig. 3, an encoding device according to a preferred embodiment of thls invention is applicable to the ADPCM system and comprises similar parts and deals with signals, ail of which will be designated by like reference numerals and symbols. The illustrated encoder 11 is similar to that illustrated in Fig. 1 except that an encoder control circuit 32 is coupled to the adaptive predictive filter 16. It should be noted here that the influence of the initial state can be removed by putting the adaptive predictive filter 16, namely, the feedback loop into a transient inactive state. For this purpose, the encoder control circuit 32 monitors the encoder predictor 16 in a manner to be described later.
In Fig. 3, the encoder 11 encodes the encoder input signal x; into the quantized predictive error signal êj supplied as the encoder output signal to the additional encoder 19. The encoder output signal is subjected to another encoding in the additional encoder 19 to be sent as the DPCM signal Uj to a transmission ~ " .
~2~70 line (not shown).
Referring to Fig. 4 afresh together with Fig.
3, the illustrated adaptive predictive filter 16 is of a Finite Impulse Response (FIR) type and comprises a delay circuit 34 comprising first through (N + l)-th taps and first through N-th delay units 361 to 36N each of which is connected between two adjacent ones of the taps, where N is representative of a natural number of, for example, ten. The quantized predictive error signal êj is successively delayed by the first through the N-th delay units 361 to 36N and appears as first through (N + l)-th tap signals Dl to DN+l through the first through the (N + l)-th taps, respectively. There~
fore, the delay circuit 34 may be named a tap signal producing circuit.
The quantized predictive error signal êj is supplied to a gain control circuit 38 for producing first through (N + l)-th coefficient signals Cl to CN+
representative of first through (N + l)-th prediction coefficients, respectively. The first through the (N + l)-th prediction coefficients are assigned to the first through the (N + l)-th taps, respectively, and are adaptively modified under control of the gain control circuit 38 in response to the quantized predictive error signal êj. As a result, the first through the (N + l)-th coeffi-cient signals Cl to CN+l are modified in accordance with the quarltized predictive error signal êj. The first through the (N + l)-th coefficient signals C
7: .
~ ;22;~7~
to C~+1 are in one-to-one correspondence to the first through the ~N + l)-th tap signals Dl to DN+l, respectively.
First through (N + 1)-th multipliers 411 to 41N+l multiply the first through the (N + 1)-th tap signals Dl to DN+l by the first through the (N + l)-th coefficient signals Cl to CN+i to calculate first through (N + l)-th products, respectively. The first through the (N + l)-th products are delivered as first through (N + l)-th product signals to an adder circuit 43, respec-tively. An i-th one of the first through the (N + l)-th product signals Pl to PN+l can be represented by:
Ci X Di.
Thus, each product can be specified by a prescribed number of values between a maximum and a minimum value, -~
both inclusive.
The adder circuit 43 sums up the first through the (N + l)-th products Pl to PN+l and produces the predictive signal Xj given by:
N+l Xj - i~ Ci x Di.
If an audio or speech signal is carried by the encoder input signal succession x, a pause or a quiescent state inevitably occurs in the encoder input signal succession x. Under the circumstances, each of the first through the (N + l)-th products becomes small during the quiescent state but the predictive signal Xj does not easily become zero.
:
~L;22%~
More particularly, let each of the tap signals Di and the coefficient signals Ci have a predetermined length of, for example, four bits. In this event, each product signal Pi can be expressed by eight bits. Each product Pi takes zero when either the tap signal Di or the coefficient signal Ci b-ecomes equal to zero.
In order to render the feedback loop into the inactive state, all of the first through the (N + l)-th product signals Pl to PN+l should be rendered zero. However, the first through the (N + l)-th product signals Pl to PN+l are not readily rendered zero at the same time during the quiescent state of the encoder input signal succession x. Thus, the feedback loop is not readily put into the inactive state.
In Fig. 4, the illustrated control circuit 32 comprises a control portion 44 supplied with the first through the (N + l)-th product signals Pl to PN+l and a threshold circuit 45 for producing a threshold signal TH representative of a threshold value. The control portion 44 monitors the first through the (N + l)-th product signals Pl to PN+l with reference to the threshold signal TH. More specifically, if the value of each product Pi is smaller than the threshold value, the control portion 44 renders the value of each product forcibly zero. Consequently, it is possible to put the adaptive predictive filter 16 into the inactive state by rendering all of the products zero during the quiescent state. For this purpose, the control portion ' , 44 supplies the adder circuit 43 with control signals CTL representative of results of monitor-ing. Production of the control signals CTL can be possible by the use of usual comparators. The control signals CTL are used in the adder circuit 43 to individually gate the first through the (N + l)-th product- signals Pl to PN+l, respec-tively. Therefore, the adder circuit 43 may be a combination of gate circuits and an adder.
As mentioned above, the adaptive predictive filter 16 renders the predictive signal x; zero during the quiescent stateO As a result, the feedback loop is transiently put into the inactive state and the influ-ence of initial state can be removed by shutting off the feedback loop in effect.
Operation of the con rol circuit 32 will be described more in detail. Let each of the coefficient signals Ci and the tap signals Di have a predetermined length of, for example, four bits. In this event, each product signal Pi can be expressed by eight bits and therefore~divisible into a higher significant part of four bits and a lower significant part of four bits.
In addition, let each product be expressed by a fixed point representation and the higher significant part alone be produced from each multiplier as each product signal Pi with the lower significant part truncated.
If Ci and Di are equal to "0011" and "0001,"
the produGt becomes equal to "0000." Therefore, the threshold value may be equal to "0000."
~/
~2~S~70 On the other hand, it will be assumed that each tap signal Di is represented in the notation of two's complement.
If the tap signal Di is equal to minus 0001, the tap signal ~i is represented as "1111" in the two's complement notation. As a result, the product is rendered into "1111" which corresponds to the decimal number -1, not zero. When each product can take a negative value, such as -1, the product signal Pi cannot become zero. As a result, the predictive signal xj also cannot become equal to zero despite the fact that each product is very small.
In this e~ent, the predictive signal xj can become equal to zero by giving the threshold circuit 45 the threshold value of "1111 n and by rendering each product signal Pi forcibly zero when each product is equal to or less than "1111." In this example, each product signal Pi is rendered zero only when each product takes ~1111 . n Thus, an absolute value of each product may be monitored by the control circuit 32~
Likewise, each product signal Pi does not become zero during the quiescent state when each product signal Pi has a too long length. In this event, the control circuit 32 may monitor an absolute value of each product and renders the absolute value zero when the absolute value is equal to or smaller than the threshold value predetermined for each product.
Anyway, the ~eedback loop, namely, the adaptive predic-tive filter 16 can be put into the inactive state ~;,`
257~
during the quiescent state by the use of the control circuit 32.
Referring to Fig. 5, a decoding device is operable in cooperation with the encoding device described in conjunction with Figs. 3 and 4 and is similar to that illustrated in Fig. 2 except t-hat the decoder predictor 26 included in the decoder 12 is coupled to a decoder control circuit 50 which may be called a decoder control circuit. In this connection, the control circuit 32 (Fig. 3) may be named an encoder control circuit. The decoder control circuit 50 is similar in structure and operation to the encoder control circuit 32 illustrated in Fig. 4 and therefore monitors each product signal Pi to render each product forcibly zero when the absolute value of each product is not greater than a threshold value which may be equal to that determined in the encoder control circuit 32.
While this invention has thus far been described in conjunction with a preferred embodiment thereof, it will readily be possible for those skilled in the art to put this invention into practice in various other manners. For example, each of the encoder and the decoder predictors 16 and 26 may be of an Infinite Impulse Response (IIR) type. In this event, each predictor 16 and 26 (Fig. 4) is supplied with a sum of the predictive signal Xj and the quantized predictive error signal ej through an additional adder, instead of the quantized predictive error signal êj. Each of the encoder and the decoder Y'' ~L222~;70 predictors may be a combination of an IIR type and an FIR type.
,,'
Claims (5)
PROPERTY OR PRIVILEGE IS CLAIMED ARE DEFINED AS FOLLOWS:
1. A method of carrying out conversion between a succession of first signals and a succession of second signals by the use of an adaptive predictor which is responsive to said second signal succession to produce a succession of predictive signals predictive of said first signals, respectively, said adaptive predictor comprising tap signal producing means having a plurality of taps assigned with prediction coefficients, respec-tively, for producing a plurality of tap signals in response to said second signal succession, respectively, coefficient signal producing means for adaptively producing a plurality of coefficient signals representative of said prediction coefficients, respectively, a plurality of multipliers responsive to said tap signals and said coefficient signals for calculating products therebetween to produce product signals representative of said products, respectively, and output means responsive to said product signals for producing said predictive signal succession, said method comprising the step of:
monitoring said products to render each of said products zero when said each product is not greater than a predetermined value.
monitoring said products to render each of said products zero when said each product is not greater than a predetermined value.
2. A method as claimed in Claim 1, said method comprising an encoding step carried out in an encoder comprising said adaptive predictor as an encoder predictor, said encoding step being for encoding the first signal (Claim 2 continued) succession given as an encoder input signal succession into the second signal succession produced as an encoder output signal succession, wherein said encoding step comprises the steps of:
subtracting the predictive signal succession of said encoder predictor from said encoder input signal succession to produce said encoder output signal succession;
delivering said encoder output signal succession to said encoder predictor to make the multipliers of said encoder predictor produce the product signals moni-tored through said monitoring step of said encoder predictor; and making said encoder predictor produce the predic-tive signal succession thereof in response to the monitored product signals given from the multipliers of said encoder predictor.
subtracting the predictive signal succession of said encoder predictor from said encoder input signal succession to produce said encoder output signal succession;
delivering said encoder output signal succession to said encoder predictor to make the multipliers of said encoder predictor produce the product signals moni-tored through said monitoring step of said encoder predictor; and making said encoder predictor produce the predic-tive signal succession thereof in response to the monitored product signals given from the multipliers of said encoder predictor.
3. A method as claimed in Claim 1, said method comprising a decoding step carried out in a decoder comprising said adaptive predictor as a decoder predictor, said decoding step being for decoding the second signal succession given as a decoder input signal succession into the first signal succession produced as a decoder output signal succession, wherein said decoding step comprises the steps of:
adding said decoder input signal succession to the predictive signal succession of said decoder predictor to produce said decoder output signal succession;
(Claim 3 continued) delivering said decoder input signal succession to said decoder predictor to make the multipliers of said decoder predictor produce the product signals moni-tored through the monitoring step of said decoder; and making said decoder predictor produce the predic-tive signal succession thereof in response to the monitored product signals given from the multipliers of said decoder predictor.
adding said decoder input signal succession to the predictive signal succession of said decoder predictor to produce said decoder output signal succession;
(Claim 3 continued) delivering said decoder input signal succession to said decoder predictor to make the multipliers of said decoder predictor produce the product signals moni-tored through the monitoring step of said decoder; and making said decoder predictor produce the predic-tive signal succession thereof in response to the monitored product signals given from the multipliers of said decoder predictor.
4. An encoder for carrying out the method of Claim 1 by encoding said first signal succession into said second signal succession, wherein said encoder comprises:
encoder monitoring means coupled to said multi-pliers for monitoring said products to render each of said products zero when said each product is not greater than said predetermined value.
encoder monitoring means coupled to said multi-pliers for monitoring said products to render each of said products zero when said each product is not greater than said predetermined value.
5. A decoder for carrying out the method of Claim 1 by decoding said second signal succession into said first signal succession, wherein said decoder comprises:
decoder monitoring means coupled to said multi-pliers for monitoring said products to render each of said product zero when said each product is not greater than said predetermined value.
decoder monitoring means coupled to said multi-pliers for monitoring said products to render each of said product zero when said each product is not greater than said predetermined value.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP139020/1983 | 1983-07-29 | ||
| JP58139020A JPS6031326A (en) | 1983-07-29 | 1983-07-29 | Method and apparatus for coding and decoding dpcm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA1222570A true CA1222570A (en) | 1987-06-02 |
Family
ID=15235599
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA000459932A Expired CA1222570A (en) | 1983-07-29 | 1984-07-27 | Predictive code conversion method capable of forcibly putting a feedback loop in an inactive state |
Country Status (3)
| Country | Link |
|---|---|
| US (1) | US4622537A (en) |
| JP (1) | JPS6031326A (en) |
| CA (1) | CA1222570A (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0649769B2 (en) * | 1985-12-14 | 1994-06-29 | 住友精化株式会社 | Emulsification method |
| US4774496A (en) * | 1986-02-28 | 1988-09-27 | American Telephone And Telegraph Company, At&T Bell Laboratories | Digital encoder and decoder synchronization in the presence of data dropouts |
| US4853969A (en) * | 1987-01-13 | 1989-08-01 | Recognition Equipment Incorporated | Quantized digital filter |
| US5218649A (en) * | 1990-05-04 | 1993-06-08 | U S West Advanced Technologies, Inc. | Image enhancement system |
| JP3225644B2 (en) * | 1992-10-31 | 2001-11-05 | ソニー株式会社 | Noise shaping circuit |
| US5592508A (en) * | 1994-09-22 | 1997-01-07 | Cooper; J. Carl | Analog signal coding and transmission apparatus and method capable of operation with multiple types of analog and digital signals |
| US8223827B2 (en) * | 2004-05-05 | 2012-07-17 | Agere Systems Inc. | Method and apparatus for generating filter tap weights and biases for signal dependent branch metric computation |
| US8633842B2 (en) * | 2009-03-31 | 2014-01-21 | Agere Systems Llc | Methods and apparatus for direct synthesis of RF signals using delta-sigma modulator |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4437087A (en) * | 1982-01-27 | 1984-03-13 | Bell Telephone Laboratories, Incorporated | Adaptive differential PCM coding |
| US4475227A (en) * | 1982-04-14 | 1984-10-02 | At&T Bell Laboratories | Adaptive prediction |
-
1983
- 1983-07-29 JP JP58139020A patent/JPS6031326A/en active Pending
-
1984
- 1984-07-27 CA CA000459932A patent/CA1222570A/en not_active Expired
- 1984-07-27 US US06/635,032 patent/US4622537A/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| JPS6031326A (en) | 1985-02-18 |
| US4622537A (en) | 1986-11-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP0288281B1 (en) | Adpcm encoding and decoding systems | |
| US4385393A (en) | Adaptive prediction differential PCM-type transmission apparatus and process with shaping of the quantization noise | |
| US4617676A (en) | Predictive communication system filtering arrangement | |
| KR930003281B1 (en) | Apparatus for trasmitting digital signals | |
| CA1148661A (en) | Adpcm system for speech or like signals | |
| CA1222570A (en) | Predictive code conversion method capable of forcibly putting a feedback loop in an inactive state | |
| US4862173A (en) | Method and circuit for carrying out forward and inverse quantization by varying a reference step size | |
| US4571737A (en) | Adaptive differential pulse code modulation decoding circuit | |
| US4626828A (en) | Adaptive predictive code conversion method of interrupting prediction and an encoder and a decoder for the method | |
| KR100514340B1 (en) | Digital data converter | |
| CA2028947C (en) | Picture coding apparatus | |
| EP0404535B1 (en) | Bandwidth compression device | |
| GB2131637A (en) | Decoder for differential pulse code modulation signals | |
| KR0138115B1 (en) | Apparatus for processing digital signal | |
| JPH0126207B2 (en) | ||
| US7098816B2 (en) | Tri-state delta codec method and system | |
| SU1753597A1 (en) | Digital voice signal code | |
| GB2214750A (en) | DPCM picture coding | |
| JPS61158218A (en) | Signal transmitter | |
| Sayood | A postfiltered DPCM scheme | |
| JP3144099B2 (en) | Adaptive dynamic range encoding or decoding apparatus | |
| Niwa et al. | A new channel bank with block companding | |
| Murakami et al. | Theoretical comparison between DPCM and transform coding regarding the robustness of coding performance for variation of picture statistics | |
| JPH0588573B2 (en) | ||
| JP2949708B2 (en) | Digital signal processor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| MKEX | Expiry |