JP2553092B2 - Echo canceller - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Object of the Invention (Industrial field of application) The present invention relates to an echo canceling device for canceling echo that causes a howling and causes hearing loss, mainly in a conference loudspeaker.

(Conventional technology and its problems) With the spread of voice conferences, it is desired to provide a loudspeaker communication device that ensures simultaneous communication and has less reverberation. An echo canceller is one that meets this demand.

FIG. 2 is a block diagram showing an example of a conventional echo canceller.

The echo canceller as shown in the figure, the receiving input terminal 1, a speaker 2, a plurality of microphones 3 ₁ 3n, transmitting an output terminal 4, a training signal source 5, estimator 6, the estimated echo path generating circuit 7 , A subtractor 8, an adder 11, and a switch A.

The signal applied to the receiving input terminal 1 is output as sound from the speaker 2, but this sound reverberates and the microphone 3 ₁
Since the collected by 3n, advance to generate a pseudo echo signal, subtracting the estimated echo signal is performed from the collected signals by the microphone 3 ₁ 3n.

Such a pseudo echo signal is generated by the estimation circuit 6 and the pseudo echo path (FIR filter) 7. That is, the estimation circuit 6 uses the impulse response sample value sequence (iτ).
(I = 1, 2, ..., τ: sample time), and the pseudo echo path generation circuit (FIR filter) 7 has (iτ) (i = 1, ..., N, N: filter length) It is composed of a convolutional computing unit with a coefficient.

The estimated echo signal generated by the pseudo echo path generating circuit 7 is subtracted from the summed signal further adder 11 is collected by the microphone 3 ₁ 3n by the subtracter 8,
It is sent to the speech output terminal 4. That is, the echo signal is subtracted from the signal sent to the transmission output terminal 4.

However, in the conference call device, it is necessary to use a plurality of microphones according to the number of talkers, but in the conventional echo canceller, as shown in FIG. 2, N microphones 3 ₁ , 3 ₂ , ... The output of n can be used only as it is.

By the way, the advantage of using multiple microphones is to secure the volume by keeping the distance between the microphone and the speaker present at the conference to be short and to suppress the quality deterioration due to reverberation. If there
The quality deteriorates due to the decrease of S / N ratio and the interference of sound waves. Therefore, each microphone output is provided with a gain adjustment circuit so that the overall output level is equivalent to one microphone, and the gain of each microphone is adjusted according to the transmission level of each microphone. It is desirable to design so that the sensitivity of a microphone that has no speech input is adjusted by adjusting the. However, if such a design is applied to a conventional echo canceller, the change in the gain adjustment circuit is perceived as a change in the echo path, and since it has a correction by adaptive operation, it cannot follow frequently the gain adjustment. was there.

For this reason, a gain adjustment circuit is provided for each of the plurality of microphones, and a storage circuit for storing the transfer function of the echo path is provided for each of the plurality of microphones, so that the gain of the gain adjustment circuit added to the output of each microphone is adjusted. Accordingly, an echo canceller that changes the output gain of the storage circuit to distinguish between the change in the gain of the gain adjustment circuit and the change in the echo path can be considered.

However, such an echo canceller has a problem in initial training. That is, such an echo canceller sends a special signal (for example, white noise) to the room for initial training prior to a call, but the time for this initial training is about 1 to 5 seconds for one echo path. Will be needed. Therefore, if initial training is performed in the echo canceller described above, it is necessary to perform initial training by the number of connectable microphones.For example, when 10 microphones are connected, it takes 10 to 50 seconds as initial training. There is a problem that it costs.

The present invention has been made in view of such problems,
It is an object of the invention to provide an echo canceller having a short initial training time.

[Configuration of the Invention] (Means for Solving the Problems) In order to achieve the above object, the echo canceller according to the present invention comprises:
A speaker, a plurality of microphones, a means for generating a pseudo echo signal related to a plurality of echo paths formed by the speaker and the microphones, and subtracting the pseudo echo signal from a signal collected by the microphones. In an echo canceller comprising a subtractor, a training signal source for generating a training signal, a connected microphone detection circuit for detecting a connected microphone among the plurality of microphones, and an echo path related to the microphone for which the connection is detected. And a training control unit for applying the training signal to perform initial training.

(Operation) In the echo canceller of the present invention, since the connected microphone is detected and the initial training is performed only on the connected microphone, the time required for the initial training can be shortened.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an echo canceller according to an embodiment of the present invention.

This echo canceller shown in the figure has a reception input end 1, a speaker 2, a transmission output end 4, a training signal source 5, a switch A, an estimation circuit 6, a pseudo echo path 7, a subtractor 8, and n microphones 9. _{1 to} 9n, gain adjusting circuits 10 _{1 to} 10n connected to the respective subsequent stages of the plurality of microphones 9 _{1 to} 9n, an adder 11 that adds these gains, and a storage circuit 12 that stores the transfer function of each echo path. _{1 to} 12n, gain circuits 13 _{1 to} 13n interlocking with the respective gain adjusting circuits 10 _{1 to} 10n, an adder 14 for adding these gains, a connection microphone detection circuit 15 for detecting a connected microphone, It includes a training control circuit 16 for performing initial training of the memory circuits 12 _{1 to} 12 n corresponding to the existing microphones.

 Next, the operation of this embodiment will be described.

First, prior to a call, the connected microphone detection circuit 15 observes the noise level in the room and determines whether or not the microphone is connected based on the presence or absence of room noise. Now suppose the microphone 9 ₁ ,
When 9 ₃ and 9 ₅ are connected, the connected microphone detection circuit 15 sends the detection result to the training control circuit 16.
The training control circuit 16 turns on the switch A to activate the training signal source 5, and the gain adjustment circuit 10 ₁
Of the impulse response of the echo path from the speaker 2 to the microphone 9 ₁ by the estimation circuit path 6 by setting the gain α ₁ of “1” to “1” and the gains α _{2 to} αn of the gain adjusting circuits 10 _{2 to} 10n to “0”. Estimated sample value sequence (iτ) (i = 1,
2, ... N) is calculated and stored in the memory circuit 12 ₁ .

Then the gain of the gain adjustment circuit 10 to "1", is set to the other "0", the sample value sequence estimates of the impulse response of the echo path from the speaker 2 to the microphone 9 ₃ (i
τ) (i = 1, 2, ... N) is obtained and stored in the storage circuit 12 ₁ .

Furthermore the gain of the gain adjustment circuit ₁₀₅ and another "1""0"
Stored in the storage circuit 12 ₅ The results performs the same processing is set to.

After the call is started, the gain αj of the gain adjusting circuit 10j varies depending on the level of the target signal Vj (t) entering each microphone 9j. At this time, the gain of the gain adjusting circuit 10j and the gain of the gain circuit 13j are equally linked. Therefore, the output of adder 14 is (I = 1, 2, ... N), which is the pseudo echo path (FIR).
The pseudo echo path (FIR filter) 7 that instantly follows changes in the gains α _{1 to} αn of the microphone is generated by giving it as a coefficient of the FIR filter of the filter 7.

Further, with respect to the fluctuation of the echo path after the start of the call, the estimated value j (iτ) detected by the estimation circuit 6 and stored in the storage circuit 12j (j = 1, ... N) is sequentially corrected. This change in the echo path is detected by all microphones 9j (j =
Although the echo signal is used when the target signal Vj (t) is not added to 1, ... n), in this case, one of the microphones (9j) has a large gain and the other gains are “0”. The estimated value j (iτ)
And rewrite the contents of the memory circuit 12j. When all target signals Vj (t) are interrupted, the operating frequency of the microphone is maintained by maintaining the gain of the microphone 9j, which has received the target signal with the highest level until then, until another microphone receives the target signal. Modify the individual echo paths accordingly.

Thus, in this embodiment, the connected microphone detection circuit 15
The microphone connected by is detected and the training control circuit 16 performs the training of the path related to the connected microphone, and the training time is shortened because the training of all the paths is not performed unlike the conventional method. You can

[Effects of the Invention] As described in detail above, according to the present invention, it is possible to provide an echo canceller with a short initial training time.

[Brief description of drawings]

FIG. 1 is a configuration block diagram of an echo canceller according to an embodiment of the present invention, and FIG. 2 is a block diagram of a conventional echo canceller. 2 ...... speaker 5 ...... training signal source 6 ...... estimation circuit 7 ...... echo path generating circuit 8 ...... subtractor 9 ₁ ~9n ...... microphone 15 ...... connecting a microphone detecting circuit 16 ...... training control circuit

 ─────────────────────────────────────────────────── --- Continuation of the front page (72) Inventor Nobuo Koizumi 1-2356 Take, Yokosuka City, Kanagawa Nihon Telegraph and Telephone Corporation, Communication Research Laboratory (72) Inventor Shigenobu Minami 70 Yanagi-cho, Saiwai-ku, Kawasaki-shi, Kanagawa Inside the Toshiba Yanagimachi Plant (72) Inventor Takashi Saeki 3-3-1 Asahigaoka, Hino-shi, Tokyo Stock Company Inside the Toshiba Hino Plant (72) Inventor Mitsunori Ogawa 1 Komukai Toshiba-cho, Kawasaki City, Kanagawa Prefecture 1 Address inside Toshiba Research Institute Co., Ltd. (56) Reference JP-A-62-72236 (JP, A) JP-A-61-227458 (JP, A)

Claims (1)

(57) [Claims] 1. A speaker, a plurality of microphones, a means for generating a pseudo echo signal relating to a plurality of echo paths formed by the speaker and the microphones, and the pseudo sound from a signal collected by the microphones. In an echo canceller comprising a subtractor for subtracting an echo signal, a training signal source for generating a training signal, a connection microphone detection circuit for detecting a connected microphone among the plurality of microphones, and a connection are detected. An echo canceller, comprising: a training control unit that applies the training signal to an echo path related to the microphone and performs initial training.