JPH02265332A - Echo canceller - Google Patents

Abstract

PURPOSE:To improve the converging speed and to keep the cancel capability by providing a cancel object value calculation means and a correction coefficient calculation means calculating the corrected coefficieut while the preset value is preserved so as to apply echo cancelling thereby reducing the effect of external disturbance or the like. CONSTITUTION:A reception input norm calculation circuit 24 calculates the norm of a received input signal string, and a retarded reception input norm calculation circuit 26 receives the reception input signal by preceding tap number to calculate its norm. Moreover, a transmission output norm calculation circuit 28 receives a transmission output signal E from a subtractor 30 to calculate the norm. Then a correction coefficient decision circuit 32 receives the norm calculated by calculation circuits 24,26,28 to calculate the correction coefficient mu. The correction coefficient mu is set while preserving the cancellation capability to be set to be reached the cancellation capability fastest. Then while the object cancellation capability is ensured even in a state of external disturbance, fast convergence is attained.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention (C) relates to an echo canceller that cancels echo while estimating the impulse response characteristic of an echo path in a hands-free telephone or the like.

[Prior Art] Conventionally, a pseudo echo for an echo is estimated by performing adaptive control to modify the impulse response characteristic so as to approximate the impulse response characteristic of the echo path, and this pseudo echo is subtracted from the actually generated echo. Various echo cancellers that cancel echoes are known. For example, as stated in Japanese Patent Application Laid-Open No. 56-153850, 1: Using the learning identification method, the transmitted signal is regarded as the difference between the true echo and the pseudo echo, and the line noise transferred from the circuit noise estimation circuit is identified. It is known to calculate a correction coefficient using the identification error transferred from the error estimation circuit, and to correct the impulse response characteristic using this correction coefficient.

[Problems to be Solved by the Invention] However, in these conventional echo cancellers, the line noise and identification error cannot be directly observed, so estimated values are used to find the optimal solution in an average sense. The coefficient has two contradictory properties; when the correction coefficient is large, the convergence speed is fast but the cancellation ability is low; when the correction coefficient is small, the convergence speed is slow but the cancellation ability is high. Since the signal is regarded as the difference between the true echo and the pseudo echo, this transmitted signal contains disturbances such as noise and the voice of the near-end speaker, and if the correction coefficient changes significantly due to such disturbances, There was a problem in that the convergence speed and cancellation ability changed greatly, and the desired properties could not be obtained.Therefore, the present invention aims to solve the above problems.
To provide an echo canceller that maintains high canceling ability while reducing the influence of disturbances and increasing convergence speed.

[Means for Solving the Problem] In order to achieve the above object, the present invention adopts the following configuration as a means for solving the problem. That is, the difference between an echo signal and a pseudo echo signal approximated thereto is calculated. and a subtracter that outputs a transmission output signal, calculates a modification coefficient from the transmission output signal and the reception input signal, performs adaptive control to modify the impulse response characteristic according to the modification coefficient, and generates the pseudo echo signal. In the echo canceller that outputs the echo to the subtracter to cancel the echo, the cancellation target value calculation is performed to gradually approach the cancellation target value for the echo cancellation to a predetermined target value based on the transmission output signal and the reception input signal. means; 1. based on the cancellation target value and the transmission output signal; This is a configuration of an echo canceller characterized by comprising: correction coefficient calculating means for calculating the correction coefficient while holding a preset value.

[Operation] The echo canceller having the above-mentioned configuration calculates a cancellation target value for echo cancellation to gradually approach a predetermined target value based on the transmission output signal and reception input signal, and calculates a correction coefficient. The calculation means calculates a correction coefficient based on the cancellation target value and the transmission output signal while holding a preset value.
performs adaptive control to modify the impulse response characteristic according to the modification coefficient and outputs a pseudo echo signal to the subtracter;
A subtracter takes the difference between the echo signal and the pseudo echo signal and outputs a transmission output signal to cancel the echo.

[Example] Embodiments of the present invention will be described in detail below based on the drawings.

FIG. 1 is a block diagram of an echo canceller according to an embodiment of the present invention. 1 is the echo canceller main body, and the echo canceller main body 1 (upper reception input terminal 2 and
A speaker output terminal 6 connected to the speaker 4, a microphone input terminal 10 connected to the microphone 8, and a transmission output terminal 1
2 is provided. In this embodiment, the microphone 8 is installed in the steering wheel of an automobile (not shown), and the speaker 4 is applied to a hands-free telephone installed in the back seat.

Further, an analog-to-digital converter 14 (hereinafter referred to as A/D converter 14) is connected to the reception input terminal 2, and after the received signal is converted into a -H digital signal, it is converted into a digital-to-analog signal again. Converter 16 (hereinafter referred to as D/A
The signal is converted into an analog signal by a converter 16) and output from the speaker output terminal 6. Then, an A/D converter 18 that converts the input signal from the microphone 8 into a digital signal is connected to the microphone input terminal 10, and a D/A converter 18 that converts the digital transmission signal to an analog signal is connected to the transmission output terminal 12. 20 are connected.

In this way, each signal is digitized within the echo canceller main body 1. In the digital signal processor (as shown in Figure 1, the received signal from the A/D converter 14) is input to the receiving input register 22 that stores the received signal for the number of taps.
The reception input register 22 parallelizes serially input reception signals. The reception input signal string x(k) parallelized by the reception input register 22 is expressed as a determinant as shown in the following equation.

X (k) ” (Xb, Xb-+, ...,
Xk-M4+) However, M is the number of taps. The number M of taps and the impulse response time of the echo path are in a proportional relationship.

Reception manual norm calculation circuit 241A calculates the norm of this reception input signal sequence x(k) using an algorithm described later.

In addition, the delayed reception manual norm calculation circuit 26 (top) receives the reception input signal a few minutes ago from the reception input register 22 and calculates its norm using an algorithm described later. The norm is calculated using the algorithm described later by receiving the transmission output signal E of the transmission output signal E.
When the microphone 8 is free of disturbances such as noise or near-end talkers, the residual echo signal is the same as above 1.

Then, the modification coefficient determination circuit 32 receives the calculated norms from the reception human power norm calculation circuit 24, the delayed reception norm calculation circuit 26, and the transmission output norm calculation circuit 28, and calculates the modification coefficient μ. As shown in the figure, 1: When the cancellation ability is constant at 20 dB in this example (upper), it has been previously determined through experiments that the disturbance and the correction coefficient μ are in an inversely proportional relationship. The cancellation ability will be higher, but the convergence speed will be slower.Then, we set the target value to 20 dB and adopt the fastest correction coefficient μ that satisfies it.
It is the value on the line of the graph.

Furthermore, as shown in FIG. 7, it has been determined in advance through experiments that when the disturbance is constant, the canceling ability and the correction coefficient μ are in direct proportion to each other.
From the relationship shown in the figure and the relationship shown in FIG. 7, the relationship between the correction coefficient μ, cancellation ability, and disturbance (above and below) can be approximated by equation (1).

μ = α゜ is the cancellation target value obtained by the algorithm described later, and OUT
is the amount of disturbance and cannot be observed directly, so in this embodiment, the norm of the transmission output signal E calculated by the transmission output norm calculation circuit 28 is used. 1, and when μ〉1 (friend
When μ=1 and μ<0.01 (9 μ20゜0
It may be set to 1.

On the other hand, the adaptive filter coefficient calculation circuit 34 (friend) receives the reception input signal sequence x(k) from the reception input register 22 and the current reception input signal sequence x (k) from the reception human power norm calculation circuit 24.
k), the correction coefficient μ from the correction coefficient determining circuit 32, and the transmission signal E from the subtracter 30, and the adaptive filter sequence W(k) is obtained by the following equation (2) using the learning identification method. Calculate by.

It is the norm of the reception input signal sequence x (k) calculated by the reception input register 22.

Then, the adaptive filter series sequence W(k) from the adaptive filter coefficient calculation circuit 34 and the reception input signal sequence x (k) from the reception input register 22 are convolved and integrated (W( k) · In this way, in this embodiment, by the learning identification method, (2
) The adaptive filter coefficients are updated using equation (3).

E=y (k)-w (k) ・x (k)-(3)
Next, the algorithm of the echo canceller of this embodiment will be explained with reference to FIGS. 2 to 5.

The received signal input to the reception input terminal 2 is converted into a digital signal by the A/O converter 14.

This digital signal is input to the reception input register 22 and converted back to an analog signal by the D/A converter 16, and the sound is emitted from the speaker 4. Part of the sound from the speaker 4 is influenced by the characteristics of the echo path and then wraps around the microphone 8.

Table 1 of reception signals input to the reception input register 22
Received input values for the number of taps are stored and output as a received input signal sequence x(k). This received input signal sequence x(k
)IL is input to the receiving human power norm calculation circuit 24,
Calculation of the norm is performed using the algorithm shown in FIG.

First, the norm PW calculated last time (n-1st time), -1
It is determined whether or not is smaller than 1/2 (IN2*number of taps/2) obtained by multiplying the square value of the current (nth) received input value IN by the number of taps (step 90). And the norm PW calculated last time.

When PWI is small, the current norm PWI is the square value of the current received input value IN: 1/2 multiplied by the number of taps.
(step 92). Here, the received input value IN
I; J is the n-th instantaneous input value, and if the input signal is a sine wave, the average value is obtained by multiplying the instantaneous input value by the number of taps, which is 1/2. As a result, approximate values can be obtained without performing complicated calculations.

On the other hand, when the previously calculated norm P W n-1 is large (friend), the current norm PWo is the previously calculated norm P
Wn-tlmi is multiplied by a predetermined value C (0<C<1) (step 94). By doing this, we are trying to gradually reduce the friend power after it reaches its peak.

Also, in the delay reception manual norm calculation circuit 26, I-1
From the instantaneous input value IN inputted from the received signal input register 22 by the number of taps, the norm PW delayed by the number of taps is calculated using the algorithm shown in FIG. Further 1:
, in the transmission output norm calculation circuit 28 (Table Subtractor 3
From the instantaneous input value IN of the current transmission output signal E from 0,
The norm PW is calculated using the algorithm shown in FIG.

Then, the respective norms l and t calculated by the reception human power norm calculation circuit 24, the delayed reception human power norm calculation circuit 26, and the transmission output norm calculation circuit 28 are transferred to the correction coefficient determination circuit 28. One peak in the correction coefficient determination circuit 28 Figures 3 to 5
Calculate the correction coefficient μ using the algorithm shown in the figure. First, calculate the current cancellation ability (step 100)
This calculation (like the algorithm shown in FIG. When the current receiving human power norm PWo is smaller than the delayed receiving input norm (step 110), the transmitting output norm II E transferred from the transmitting output norm calculation circuit 28 is is divided by the current receiving human power norm PW to calculate the current cancellation ability (=transmission output norm/receiving human power norm) (step 120).

On the other hand, when the current reception input norm PWo is greater than or equal to the delayed reception input norm (step 130), the current reception input norm PWo is replaced with the delayed reception input norm (step 130), and the transmission output norm It
Calculate the current cancellation capability by dividing E I+ 2 by the replaced current receiver power norm (step 120).

In this way, when calculating the current cancellation ability, it is based on the norms of both the transmitting output signal E and the receiving input, but until the echo from the receiving input returns to the transmitting output via the echo path, Since there is a delay, when the power rises, it is judged that the canceling ability is better than the current canceling ability.Therefore, compare the norm of the reception input delayed by the number of taps with the current reception human power norm, and select the smaller one. This is prevented by adopting .

Next, the cancellation target value C8o is updated (step 200). As shown in the algorithm of FIG. 5, this update is performed by first canceling the target value C8.

is larger than the current cancellation ability +3 dB (step 210). Add a predetermined value a to the cancellation target value C3o-+ and set it as a new cancellation target value CS, (=CSn-10a) (step 220).On the other hand, the current cancellation ability +3dB When the above is satisfied, a new cancellation target value C8o is calculated using the following formula.

C8,=C8,-,+1/b. (Current cancellation ability +3 dB C5n-+) Here, b is a time constant that brings C8o closer to the current cancellation ability +3 dB. In this way, the cancellation target value CS is updated so as to gradually approach the target value, with the current cancellation ability +3 dB as the target value.

After updating the cancellation target value C8o in this way, the correction coefficient μ is calculated based on the above equation (1). Here, since the amount of disturbance OUT cannot be directly observed, in this embodiment (above) the transmission output norm calculation circuit 28
The transmission output norm It E I+ 2 calculated by is used.

If the transmission output signal E is used as the amount of disturbance OUT, the echo of 1 will be considered as a disturbance at the beginning of learning, and the correction coefficient μ will be lowered more than necessary. So, current cancel ability +
I move it so that it follows 3 dB to reach the target value. This makes it possible to reduce the convergence time required during initial learning.

Next, this calculated correction coefficient μ(1) is transferred to the adaptive filter coefficient calculation circuit 34, and the adaptive filter coefficient calculation port j834 calculates the correction coefficient μ(1), the transmission output signal E, and the reception input signal sequence x(k). , receiver human power norm 1lx(k)II2
A new adaptive filter sequence w(k+1) is calculated from the adaptive filter sequence w(k) using the above equation (2).

In the learning identification method, the norm of received input is
) Due to the integral effect of 11', the operation may become unstable at the rise of 1 mi power.

On the other hand, the algorithm shown in FIG. 2 makes the operation stable even when the power is rising.

Then, this new adaptive filter sequence w(k+1) is transferred to the convolution operator 36, and the convolution operator 36
Then, the received input signal sequence x (k) and the adaptive filter sequence w (k+1) are convolved and integrated (W (k + 1)
・x(k)) to create a pseudo echo signal. Then, the pseudo echo signal is transferred to the subtracter 30, and the subtracter 30 subtracts the pseudo echo signal from the microphone input signal y(k) as shown in equation (3) above.
It is output as a transmission output signal E. This transmission output signal El
The signal is converted into an analog signal by the D/A converter 20 and transmitted via the transmission output terminal 12.

As described above, the echo canceller 1 of this embodiment calculates the respective norms of the received input signal sequence X(k), the delayed received input signal, and the transmitted output signal E using the algorithm shown in FIG. Then, the modification coefficient determination circuit 32 determines the modification coefficient μ using the algorithm shown in FIGS. 3 to 5. By determining this correction coefficient μ, first, the current cancellation ability +
With 3 dB as a target value, the cancellation target value C8n is gradually changed to reach this target value. Then, from this cancellation target value C8o and the norm of the transmission output signal estimated to be the disturbance, the correction coefficient μ is determined while maintaining the cancellation ability. With this correction coefficient μ, (2)
, performs the operation expressed by equation (3), performs an adaptive filter operation, and creates a pseudo echo signal. Then, a transmission output is created by subtracting the pseudo echo signal from the microphone input signal.

Therefore, the cancellation target value C8o is made variable so that the cancellation target value C81 gradually approaches the current cancellation ability +3 dB. Further, the modification coefficient μ is set with the cancellation ability retained, and the modification coefficient μ is determined so that the cancellation ability is reached as quickly as possible. Therefore, in an environment where it is possible to cancel more than the cancellation target value C8, 1. t, the correction coefficient μ is changed so that the convergence speeds up accordingly, and even if the cancellation ability does not reach the cancellation target value C8n, the cancellation target value C8o is reached. This makes it possible to quickly converge while ensuring the target cancellation ability even in the presence of disturbances.

As described above, the present invention is not equally limited to these embodiments, and may be implemented in various forms without departing from the gist of the present invention.

[Effects of the Invention] As detailed above, the echo canceller 1 of the present invention has a predetermined canceling ability, and even in a state where there is a disturbance, fast convergence is achieved while ensuring the canceling ability. It has the effect of being able to.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an echo canceller according to an embodiment of the present invention. FIG. 2 is a flowchart showing the norm calculation algorithm of this embodiment. FIG. 3 is a flowchart showing the correction coefficient determination algorithm of this embodiment. FIG. 4 is a flowchart showing the cancellation ability calculation algorithm of this embodiment. FIG. 5 is a flowchart showing the target updating algorithm of this embodiment, FIG. 6 is a graph showing the relationship between the correction coefficient and disturbance when a predetermined canceling ability is achieved in this embodiment, and FIG. 3 is a graph showing the relationship between a correction coefficient and cancellation ability when is constant. 1...Echo canceller body 22...Reception input register 24...Reception manual norm calculation circuit 26...Delayed reception input norm calculation circuit 28...Transmission output norm calculation circuit 32...Modification coefficient determination circuit 34...Adaptive filter coefficient calculation circuit 36...Convolution calculator

Claims (1)

[Scope of Claims] A subtracter that takes the difference between an echo signal and a pseudo echo signal approximated thereto and outputs a transmission output signal, and calculates a correction coefficient from the transmission output signal and the reception input signal, In an echo canceller that performs adaptive control to modify the impulse response characteristic according to the modification coefficient and outputs the pseudo echo signal to the subtracter to cancel the echo, the echo cancellation target value is set to the transmission output signal and Cancellation target value calculating means for calculating a value gradually approaching a predetermined target value based on a received input signal; Calculating the correction coefficient while holding a preset value based on the cancellation target value and the transmitted output signal. An echo canceller comprising: a correction coefficient calculation means for calculating a correction coefficient;