JPS6214144B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention has four separate upstream and downstream signal transmission lines.
Regarding an echo canceler that attempts to remove this leakage component (echo) when the transmitted signal leaks to the receiving side when transmitting signals in both directions simultaneously in a wire system or a two-wire signal transmission path that uses both up and down channels. It is something. When audio and digital data are transmitted using a four-wire or two-wire transmission line, echoes may occur due to line-to-line admittance or mismatch in a hybrid circuit. As a conventional and well-known echo canceller, there is a method as shown in FIG. In FIG. 1, 1 is an impulse transmitter, 2 is a signal transmitter, 3 is a hybrid circuit, 4 is a shift register, 5 is an impulse response, and 6 is a transversal filter. That is, before starting signal transmission, an impulse is sent out from the impulse transmitter 1 in advance, and the leakage component to the receiving side is captured and stored in a memory for a certain period of time from the impulse sending time. The leakage component stored in the memory is the impulse response 5 of this signal transmission system and provides the transfer function of the system.
In general, the response of a system is given by the convolution integral of the input signal and the impulse response of the system, so if we find the convolution integral of the impulse response stored in the memory and the transmitted signal, we can calculate the response of the system to the transmitted signal, that is, the receiving side This means that the leakage components have been found. Therefore, by subtracting this on the receiving side, the leakage component of the transmitted signal in the received signal will be removed. The actual convolution integral is calculated by a transversal filter 6 using the impulse response 5 as a coefficient, as shown in FIG. However, in this method, (1) it is necessary to send out an accurate impulse before transmitting the signal, but it is impossible to send out an accurate impulse, and the impulse response obtained has to be an approximate pulse. becomes that much more inaccurate. (2) The impulse response obtained prior to signal transmission must be free from noise or signals from the other party. (3) High-speed multipliers are required to complete the convolution operation within one sample period, and a large number of multipliers are required. (4) Therefore, it is necessary to cut off the impulse response midway, and the calculation error caused by the cut off part cannot be ignored. (5) Furthermore, in convolution integration, since a large number of multiplication results are accumulated, errors overlap considerably and the calculation bit length must be large. (6) Unable to follow changes in line conditions. There are many drawbacks such as. An object of the present invention is to solve the above-mentioned drawbacks, and to extract the leakage component itself on the receiving side due to the transmitted signal from the received signal occurring on the receiving side by using the statistical properties of the signal, and to use this extracted component. The present invention provides an echo canceller that removes leakage components of transmitted signals. In order to achieve the above object, the present invention provides an echo canceler that removes leakage components of a transmitted signal to the receiving side in the transmission of analog information including digital information and voice, in which a sample of the signal occurring on the receiving side is fixed. It has means for collecting numbers, and an arithmetic circuit that calculates the arithmetic mean of the sample values, and extracts the output value of the arithmetic circuit as a leakage component of the transmitted signal to the receiving side, and extracts the output value of the arithmetic circuit as a leakage component of the transmitted signal to the receiving side. The present invention is characterized in that the leakage component of the transmitted signal to the receiving side is removed by subtracting the extracted leakage component from the signal. Embodiments of the present invention will be described below with reference to the drawings.
FIG. 2 shows a principle diagram of an echo suppressor according to the invention. In Fig. 2, 7 is a transmitter, 8 is a hybrid circuit, 9 is a switch A, 10, 11, 1
Reference numeral 2 indicates an average value calculation circuit, 13, 14, and 15 memories, and 16 a switch B, respectively. In short, this method eliminates the leakage components on the receiving side by generating the leakage components themselves on the receiving side due to the transmitted signal, instead of synthesizing them from the impulse responses shown in Figure 1, which is the conventional method. The method is characterized in that the extracted components are extracted from the received signal using the statistical properties of the signal, and the extracted components are used to remove the leakage components of the transmitted signal. Now _, there are _N signals a ₁ , _{a 2} , . Meanwhile, at the same time, the first switch of switch A9 and switch B16 is closed by transmitting the _a1 signal. When the first switch of switch A9 is closed, the past data is read from the memory 13, and the average value calculation circuit 10 compares the echo sent from the hybrid circuit 8 side, the signal component and noise sent from the other party, and the The average value with past data is calculated. This average value gives a leakage component (echo) of the signal _a1 to the receiving side. Next, the contents of the memory 13 are updated, and the obtained average value is subtracted from the input signal passed through the hybrid circuit 8 by a subtracter via the first switch B16, thereby canceling the echo. The same operation is performed for the other signals a ₂ , a ₃ . . . a _N , and echoes are canceled. Next, the principle of the present invention will be explained. For simplicity, consider transmission of a signal whose distribution has an average of 0 (for example, when ±1 is transmitted with equal probability). If not, some modification is required. Now,
There are N signals a ₁ , a ₂ , ..., a _N as transmission signals, and now consider extracting the leakage component (echo) to the receiving side from the receiving side signal when a ₁ is transmitted. . When a ₁ is transmitted, the echo that appears on the receiving side is a ₁ ′, and the signal components occurring on the receiving side are the signal component b sent from the other party, the noise n, and the signal of the signal sent in the past. This can be said to be a mixture of residual leakage component C. That is, it is given by y=a ₁ '+b+n+C (1). In equation (1), a' ₁ is the echo that appears on the receiving side when a ₁ is transmitted, so if we consider only the transmission of a ₁ , the echo that appears on the receiving side is always a' ₁ . Therefore, its arithmetic mean Σ
a' ₁ /k is equal to a' ₁ . Next, since the signal b transmitted from the other party and generated on the receiving side is obtained from a transmitted signal whose distribution has an average of 0, the average of its distribution is also 0. That is, If the mean of the distribution of the noise n is also 0, then It is. Furthermore, the residual portion C of the echoes of the signals a ₁ to an transmitted in the past is also the transmitted signal a ₁ whose distribution average is 0.
Since it is given by ~an, its average is also becomes. Therefore, b, n, and c in equation (1) converge to 0 by the arithmetic mean, and equation (1) becomes Therefore, by calculating the arithmetic mean of y for a sufficiently large K, it is possible to extract the leakage component (echo) of the transmitted signal _a1 to the receiving side. If the same thing is done for the other signals a ₂ , a ₃ , . . . a _N , it is possible to obtain the leakage components of all the transmitted signals. By constantly updating these K samples y, it is possible to extract echoes that follow fluctuations in the line. Some embodiments of the present invention will be described based on the above principle. FIGS. 3, 4, and 5 are examples in which there is no inter-symbol interference of the transmitted signal, and FIGS. 6 and 7 are examples in which there is inter-symbol interference in the transmitted signal. (1) Case without inter-symbol interference When there is no inter-symbol interference, the problem is simple and can be said to be the same as the principle in Figure 2. Figure 3 shows a case where the average value calculation circuit of the principle diagram in Figure 2 is shared. An example is shown and works as follows. The average of the distribution of the transmission signals a ₁ , a ₂ , . . . a _N is 0, and the memory 13 corresponds to a ₁ and the memory 15 corresponds to a _o . First, do the same as the principle diagram in Figure 2.
When a ₁ is transmitted, the switch 9 is connected to the memory 13, reads past data from the memory 13, extracts the average value, that is, the echo of the transmitted signal a ₁ by the average value calculation circuit 10, and converts the echo through the subtracter. Cancel. At this time, the data in the memory 3 is updated at the same time. The same thing is true for a ₂ ,
This is also done for a ₃ , ...a _o , and the echo is canceled. Since the average value calculation circuit 10 does not need to operate simultaneously for each transmission signal,
It can be shared for each transmission signal. That is, the embodiment shown in FIG. 3 uses the average value calculation circuits 10 to 12 in the principle diagram (FIG. 2) in common. The reason why they can be shared is that although the average value calculation circuit in the principle diagram is provided for each transmitted signal, these circuits have the same structure, and moreover, many types of signals are not transmitted at the same time. In FIG. 4, the switch 9 in FIG. 3 corresponds to memory address selection, and transmission signals a ₁ , a ₂ , . . .
..., a _o correspond to memory address specification, the switch becomes unnecessary. A configuration that further reduces memory can also be considered. That is, focusing on one of the transmitted signals,
Only the leakage components of this transmission signal are extracted, and the leakage components of other signals are converted from the extracted leakage components of the transmission signal.
In this way, the samples of signal components on the receiving side that should be retained and the extracted leakage components are
All that is required is the transmission signal of interest. At this time, when transmitting other signals, the average value calculation circuit 10 may convert the leakage component to the receiving side so as to correspond to the leakage component of the transmission signal of interest. That is, in FIG. 4, the memories 13 and 14 are further unified and constituted by one large memory, and the switch function is replaced with a memory address designation function. The memory map in FIG. 4 is as shown in Table 1, for example, and the switch function is replaced by selecting addresses a ₁ -a _o .

[Table] Now, Fig. 5 shows that a ₁ , a ₂ , ..., a
This is an example of a configuration when a multilevel signal of _o is selected.
When transmitting a signal other than the transmission signal of interest, the switch 9 selects one of the conversion levels b ₁ to b _o , the multiplier adjusts the level, and the average value is calculated. Further, the switch 16 is switched in response to the transmission of a signal other than the transmission signal of interest, and after adjusting the level of the leakage component of the corresponding transmission signal, canceling the echo is performed. Of course, if the average value is not calculated when transmitting signals other than the transmission signal of interest, the switch 9 is unnecessary. If the number of transmission signals of interest is not one but a plurality of transmission signals, the memory may be set to correspond to the number of transmission signals. That is, the embodiment shown in FIG. 5 is an example in which the transmission signal is limited to a multilevel signal. At this time, for example, a typical signal is
When a ₁ is selected, other signals can be converted by multiplying this a ₁ by a constant. If this conversion is performed on the receiving side, all after conversion (output of the multiplier that multiplies b ₁ to b _o ) can be processed as if a ₁ was transmitted, regardless of the transmitted signal. That is, in the principle diagram (FIG. 2), the average value calculation circuit and memory prepared for each transmission signal need only be provided for one type of signal _a1 .
However, since the echo is only obtained for a ₁ at this point, inverse conversion is required to obtain echoes of other signals. Furthermore, in FIGS. 3, 4, and 5, if there is a delay in the leakage component of the transmitted signal to the receiving side, the switch in FIGS. Unless the addressing is made to guarantee the delay, the memory needs to be increased by the amount corresponding to the delay. (2) When there is intersymbol interference Intersymbol interference means that an echo of a signal transmitted at a certain time and an echo transmitted at another time exist at the same time. In this case, even after the transmission of the signal being canceled ends, the echo of that signal still continues, so it is necessary to continue canceling the echo of that signal even while the next signal is being transmitted. Therefore, this configuration corresponds to the case with the delay described in the previous section 1. An example of its configuration is shown in FIG. FIG. 6 shows the configuration of FIG. 4 applied to this case. Similarly, a similar configuration can be considered for FIGS. 3 and 5. The switch 9 is turned on every time a signal is sent, and continues to be turned on until intersymbol interference disappears. That is, FIG. 6 is an extension of the embodiment of FIG. 4 to the case where there is inter-symbol interference. Now, the fact that there is intersymbol interference means that the echoes of signals transmitted in the past remain even at this point in time, so we must preserve the echoes going back to the past until the intersymbol interference disappears. It won't happen. If the number is m,
m memories are required. The memory 13 shown in FIG. 6 includes m memories. Since the current echo is given as the sum of the echoes of the past m transmitted signals, the switch 9 and the cumulative adder Σ are used to calculate the sum.
is available. For example, if the transmitted signal is Assuming that, the current echo portion a′ ^m _{1 of m signals a 1 transmitted} in the past is the memory a ₁ that stores echoes of m signals transmitted in the past.
is read from the address corresponding to . Similarly,
The echo corresponding to the signal _a3 transmitted m-1 times in the past is read from the address corresponding to _a3 of the memory that stores the echoes of the signal transmitted m-1 times in the past. The echoes corresponding to the above signal sequence (a ₁ a ₃ a ₇ . . . a ₂ ) can be obtained by reading out the signals in the same manner and adding them. Another configuration is a method of calculating leakage components for all combinations of signals occurring in a range thought to be affected by intersymbol interference. 7th
The figure shows an example of its configuration. Now transmit signals a ₁ , a ₂ ,
..., _ao , and if the range of intersymbol interference is m in number of signals, then there are m ⁿ types of transmission signal patterns that occur within this range. Therefore, the memory needs to store data for leakage components corresponding to m ⁿ transmission signal patterns. Now, m signals transmitted in the past have been written in the buffer register 17, and since echoes corresponding to the signals in the register 17 appear on the receiving side, the average value calculation circuit 10 stores them in the memory. The leakage component can be canceled by obtaining the echo and subtracting it using the data within. That is, FIG. 7 is a system in which the memory 13 of FIG. 6 is unified, and the cumulative addition of echoes of the past m signals is stored in the memory. For example, given the above signal sequence (a ₁ a ₃ a ₇ ... a ₂ ), if you memorize the current echo for that combination, when this combination is obtained, you can retrieve it from the corresponding memory address. , you can get the current echo just by reading its contents. There are n _n combinations of the signal strings. In the above embodiments, the case where the same type of signal is used as the transmission signal has been shown, but when a plurality of different signals are used for signal transmission, (1) Collect a fixed number of samples of the signal that is occurring, extract the leakage component corresponding to the transmitted signal from the set of samples using the law of large numbers, and extract one of the extracted leakage components. Or, by preparing multiple units, selecting one of them each time the corresponding transmission signal is sent out, and finding the difference between it and the signal occurring on the receiving side, the receiving side of the transmitted signal (ii) collect a certain number of sets of signal samples that occur on the receiving side only for one or more different transmitted signals; Even when a transmission signal other than the above is transmitted, the signal occurring on the receiving side is collected in a fixed number of sets, including samples converted to the case of sending the transmission signal, and corresponds to the transmission signal. Extract the leakage components using the law of large numbers, prepare one or more of the extracted leakage components, and when the transmission signal is sent, one of the extracted leakage components. If a transmission signal other than the selected transmission signal is transmitted, the transmission signal is determined by calculating the difference between the extracted leakage component selection conversion and the signal occurring on the receiving side. The echo is canceled by removing the leakage components to the receiving side. In addition, when using transmission signals that occur within an arbitrary fixed period of time, a fixed number of sets of samples of signals occurring on the receiving side are collected for each combination pattern of signals and information to be transmitted, and these samples are Applying the law of large numbers to the set of
Extract the leakage component to the receiving side corresponding to the combination pattern, and in response to the appearance of the combination pattern,
The echo is canceled by removing the leakage component of the transmitted signal by taking the difference between the extracted leakage component and the signal occurring on the receiving side. Further, in the above embodiments, a two-wire system using a hybrid circuit has been described, but it goes without saying that the present invention can also be applied to echoes occurring in a four-wire system or leakage from other lines. As explained above, the present invention has the following effects. (1) Since the leakage components of the transmitted signal are extracted from the signals occurring on the receiving side, there is no need to send out impulses prior to signal transmission. (2) Therefore, noise and signal contamination from the other party will not be a decisive hindrance to the cancel operation. (3) Since the leakage component itself of the transmitted signal is extracted and used to cancel the leakage component, convolution is not required and the hardware configuration is simplified. (4) Since no convolution is required, there is no accumulation of multiplication errors, and the operation bit length can be short. (5) Since leakage components of the transmitted signal are constantly extracted and updated, it is possible to follow fluctuations in the line. (6) In principle, nearly 100% cancellation is possible because the leakage components of the transmitted signal are used to remove echoes.

[Brief explanation of the drawing]

Fig. 1 is a conventional echo canceller, Fig. 2 is a principle diagram of an echo canceller according to the present invention, Fig. 3,
4 and 5 are embodiments according to the present invention in which there is no inter-symbol interference, and FIGS. 6 and 7 are embodiments in which there is inter-symbol interference according to the present invention. 1... Impulse transmitter, 2... Signal transmitter,
3...Hybrid circuit, 4...Shift register, 5...Impulse response, 6...Transversal filter, 7...Transmitter, 8...Hybrid circuit, 9...Switch A, 10, 11, 12...
...Average value calculation circuit, 13, 14, 15...Memory, 16...Switch B, 17...Buffer register.

Claims (1)

[Claims] 1. Means for collecting a certain number of samples of signals occurring on the receiving side in an echo canceler that removes leakage components of transmitted signals to the receiving side in analog information transmission including digital information and audio. and an arithmetic circuit that calculates the arithmetic mean of the plurality of sample values, and extracts the output value of the arithmetic circuit as a leakage component of the transmitted signal to the receiving side, and extracts the signal occurring on the receiving side. An echo canceller characterized in that a leakage component of a transmission signal to a receiving side is removed by subtracting the extracted leakage component from . 2. The means for collecting a fixed number of samples of signals occurring on the receiving side collects a fixed number of samples in sets corresponding to the same type of transmitted signals, as set forth in claim 1. Echo canceller. 3. The means for collecting a fixed number of samples of signals occurring on the receiving side is characterized in that it collects a fixed number of sets of signals occurring on the receiving side for one or more different transmitted signals. An echo canceller according to claim 1. 4. The above means for collecting a certain number of samples of signals occurring on the receiving side is a means for collecting a certain number of samples of signals occurring on the receiving side for transmission signals other than the reference transmission signal. 2. The echo canceller according to claim 1, wherein a fixed number of samples of the converted signal are collected in sets. 5 The means for collecting a fixed number of samples of signals occurring on the receiving side is to collect a fixed number of samples of signals occurring on the receiving side for each combination pattern of transmission signals to be transmitted that occurs within a fixed time. The echo canceller according to claim 1, characterized in that the echo canceller is collected in sets.