JP2004080647A - Echo canceller and telephone conversation processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide echo cancelling of high quality in auditory perception. <P>SOLUTION: An echo canceller comprises an adaptive filter 11 which estimates echo and outputs it as a first pseudo echo signal s1, and a fixed filter 12 which estimates an echo according to the factor transferred from the adaptive filter 11 and outputs a second pseudo echo signal s2. A switching control unit 15 controls a selector 14 to output the signal which is echo-cancelled by the first pseudo echo signal s1 outputted from the adaptive filter 11 instead of the signal which is echo-canceled by the second pseudo echo signal s2 outputted from the fixed filter 12, if the converging state of the adaptive filter 11 is stable or an echo suppression amount ERLE improves in the signal echo-cancelled by the second pseudo echo signal s2 outputted from the adaptive filter 11. <P>COPYRIGHT: (C)2004,JPO

Description

そして、適当なフィルタを用いて送話信号ｉからカーノイズ成分を抽出し、カーノイズのレベルが所定のレベル以上かどうかを調べ（ステップ２１０）、所定のレベル以上であれば、第１送話信号ｅ１を出力するようにセレクタ１４を制御する（ステップ２１４）。これは、カーノイズのレベルがあるレベル以上大きい場合、前述のように固定フィルタ１２による第２送話信号ｅ２よりも、適応フィルタ１１の生成する第１送話信号ｅ１を選択した方が聴覚上高品質なエコーキャンセルを行うことができる可能性が高いからである。ここで、このように第１送話信号ｅ１に選択を切り替えるしきい値となるカーノイズの所定のレベルは、たとえば、予め実験的に求めて設定しておく。
【００２７】
一方、もし、カーノイズのレベルが所定のレベル以上でなければ、その時点で、適応フィルタ１１の収束状態が安定しているかどうかを、係数誤差ＣＥが所定レベル以下であるかどうかや、適応フィルタ１１によるエコー抑圧量ＥＲＬＥ　の時間変動が所定レベルより小さいかどうかより判定し（ステップ２１２）、適応フィルタ１１の収束状態が安定していれば、第１送話信号ｅ１を出力するようにセレクタ１４を制御する（ステップ２１４）。これは適応フィルタ１１の収束状態が安定していれば、第１送話信号ｅ１において適応フィルタ１１により聴覚上高品質なエコーキャンセルが実現されていると考えることができるからである。
【００２８】
ここで、係数誤差ＣＥは、第１送話信号ｅ１のレベルが大きいほど大きくなる特性を有し、適応フィルタ１１によるエコー抑圧量ＥＲＬＥと、たとえば図３に示すような一定の相関を有する。
次に、もしカーノイズのレベルが所定のレベル以上でなく、かつ、適応フィルタ１１の収束状態が安定してなければ、適応フィルタ１１によるエコー抑圧量ＥＲＬＥが、時間的に増加傾向にあるかどうかを調べ（ステップ２１６）、増加傾向にある場合には、第１送話信号ｅ１を出力するようにセレクタ１４を制御し（ステップ２１４）、増加傾向にない場合には第２送話信号ｅ２を出力するようにセレクタ１４を制御する（ステップ２１８）。これは適応フィルタ１１によるエコー抑圧量ＥＲＬＥが増加していれば、適応フィルタ１１の過去の係数が適用される固定フィルタ１２による第２送話信号ｅ２よりも、第１送話信号ｅ１の方が現在においてまたは極めて近い将来において、より聴覚上高品質なエコーキャンセルが実現されると考えることができるからである。
【００２９】
次に、図４に、切替制御部１５が行う係数転送の制御の処理の手順を示す。
図示するように、この処理では、電話インタフェース部４から通知される回線接続状態から、回線接続状態が回線が接続した状態に変化したかどうかを調べ（ステップ４０２）、回線が接続した状態に変化したならば、適応フィルタ１１の係数を固定フィルタ１２に転送する（ステップ４１０）。これは、発着信開始時から回線接続までの間は、近端のユーザの発話は行われず非ダブルトーク状態となり、したがって、回線接続時には適応フィルタ１１に適切な係数が設定されているであろうことが期待できるからである。
【００３０】
また、この処理では、ダブルトーク検出部１３がダブルトーク状態を検出中であるかどうかを調べ（ステップ４０４）、ダブルトーク状態を検出中でなく、適応フィルタ１１のＥＲＬＥが固定フィルタ１２のＥＲＬＥより良好であれば（ステップ４０６）、適応フィルタ１１の係数を固定フィルタ１２に転送する（ステップ４１０）。
【００３１】
ただし、以上のステップ４０６では、ＥＲＬＥと共に係数誤差ＣＥを考慮して、適応フィルタ１１の係数を固定フィルタ１２に転送するかどうかを判定するようにしてもよい。すなわち、たとえば、適応フィルタ１１のＥＲＬＥが固定フィルタ１２のＥＲＬＥより良好であって、係数誤差ＣＥが所定レベルより小さい場合にのみ適応フィルタ１１の係数を固定フィルタ１２に転送するなどしたり、適応フィルタ１１のＥＲＬＥが固定フィルタ１２のＥＲＬＥより良好である程度と係数誤差ＣＥの小ささの程度を乗じた値が所定レベル以上であるときに、適応フィルタ１１の係数を固定フィルタ１２に転送するなどするようにしてもよい。このようにＥＲＬＥと共に係数誤差ＣＥを考慮することにより、適応フィルタ１１の安定度をも加味して、より適切に、真に現在の適応フィルタ１１の係数が、現在の固定フィルタ１２の係数よりもエコーキャンセルに適したものであるかどうかを判定することができるようになる。なお、このような係数誤差ＣＥを考慮した係数の転送制御は、図６に示した従来のダブルフィルタアルゴリズムのエコーキャンセラにおける適応フィルタ１１から固定フィルタ１２への係数の転送制御に同様に適用可能である。
【００３２】
一方、ダブルトーク検出部１３がダブルトーク状態を検出中の場合には、第１送話信号ｅ１と第２送話信号ｅ２のレベルが所定レベル以上（たとえば、２倍以上）異なるかどうかを調べ（ステップ４１２）、異なる場合には、逆に固定フィルタ１２の係数を適応フィルタ１１に転送し（ステップ４１４）、フィルタ部１１２に設定する。また、固定フィルタ１２のＥＲＬＥが適応フィルタ１１のＥＲＬＥがより数倍（たとえば、４〜８倍）以上良好である場合にも（ステップ４０８）、固定フィルタ１２の係数を適応フィルタ１１に転送し（ステップ４１４）、フィルタ部１１２に設定する。
【００３３】
また、ダブルトーク検出部１３がダブルトーク状態を検出中でない場合であっても、固定フィルタ１２のＥＲＬＥが適応フィルタ１１のＥＲＬＥがより数倍（たとえば、４〜８倍）以上良好であれば（ステップ４０８）、固定フィルタ１２の係数を適応フィルタ１１に転送し（ステップ４１４）、フィルタ部１１２に設定する。このような場合には、適応フィルタ１１の係数は比較的大きく誤っており、固定フィルタ１２の係数を適応フィルタ１１に戻して、その係数から係数更新動作を行わせた方が、適当な係数への到達をより速やかに行うことができると考えられるからである。
【００３４】
以上、本発明の実施形態について説明した。
ところで、以上では、車載用のハンズフリー通話装置への適用を例にとり本実施形態を説明したが、本実施形態に係るエコーキャンセラは、任意の音声通信装置に適用可能である。また、音声通信装置以外の任意の音声処理装置１におけるエコーキャンセルに、本実施形態に係るエコーキャンセラは適用可能である。たとえば、図５ａに示すようなマイクロフォン３からの音声認識装置５１への音声入力への、車載オーディオ装置やナビゲーション装置５２のスピーカ２からの出力音声の回り込みのキャンセルや、カラオケ装置５３における、マイクロフォン３からの音声入力への、スピーカ２から出力されるカラオケ演奏音楽の回り込みなどにも同様に適用可能である。ただし、車載以外の環境で使用される音声処理装置１へ適用する場合には、図２ステップ２１０ではカーノイズに代えて当該環境下で生じるノイズを、セレクタ１４の出力の切替に用いるようにする。
【００３５】
【発明の効果】
以上のように、本発明によれば、より聴覚上高品質なエコーキャンセルを行うことができるエコーキャンセラを提供することができる。
【図面の簡単な説明】
【図１】本発明の実施形態に係るハンズフリー通話装置の構成を示すブロック図である。
【図２】本発明の実施形態に係るセレクタの出力切替制御の手順を示すフローチャートである。
【図３】本発明の実施形態で用いる係数誤差とＥＲＬＥの相関を示す図である。
【図４】本発明の実施形態に係る係数転送制御の手順を示すフローチャートである。
【図５】本発明の実施形態に係るエコーキャンセラの他の応用を示す図である。
【図６】従来のエコーキャンセラの構成を示すブロック図である。
【符号の説明】
１：音声処理装置、２：スピーカ、３：マイクロフォン、４：電話機インタフェース装置、５：携帯電話機、１１、６１：適応フィルタ、１２、６２：固定フィルタ、１３、６３：ダブルトーク検出部、１４：セレクタ、１５：切替制御部、１６、６４：第１加算器、１７、６５：第２加算器、１８：アンプ部、６６：ダブルフィルタ制御部、１１２：フィルタ部。[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an echo canceller suitable for use in a noisy environment.
[0002]
[Prior art]
The echo canceller is applied to, for example, a communication device that performs voice communication. In such a communication device, the echo canceller performs an operation of canceling a component (echo) of a received voice output from a speaker that goes around to a microphone from a transmitted voice.
[0003]
Here, FIG. 6 shows a configuration of a conventional echo canceller.
The illustrated echo canceller is an echo canceller to which an algorithm called a double filter algorithm is applied, and includes an adaptive filter 61 (sometimes called a “background filter”) and a fixed filter 62 (sometimes called a “foreground filter”). ), A double talk detector 63, a double filter controller 66, a first adder 64, and a second adder 65.
[0004]
The operation of the present echo canceller will be described by taking as an example a case where the present invention is applied to a communication device that performs voice communication. The adaptive filter 61 estimates a echo signal from a line reception signal o and outputs it as a first pseudo echo signal s1. From the error signal e1 between the input transmission signal i and the first pseudo echo signal s1 obtained by the first adder 64, the impulse response of the echo transmission system of the acoustic echo into the input transmission signal i of the line reception signal o. And a coefficient updating unit for sequentially updating the filter coefficients to coefficients realizing the identified impulse response.
[0005]
The fixed filter 62 is a filter to which the coefficient of the adaptive filter 61 is transferred. The fixed filter 62 estimates an echo signal from the line reception signal o and outputs it as a second pseudo echo signal s2. Then, the second adder 65 subtracts the second pseudo echo signal s2 from the input transmission signal i to obtain a line transmission signal e2.
[0006]
Here, the transfer of the coefficients from the adaptive filter 61 to the fixed filter 62 is performed when it is determined that the impulse response of the adaptive filter 61 matches the actual echo transmission system more than the impulse response of the fixed filter 62. Done. Conventionally, this determination is based on the fact that the power of the error signal (residual echo) e1 between the input transmission signal i and the first pseudo echo signal s1 is equal to the error signal e2 (the error signal e2 () between the input transmission signal i and the second pseudo echo signal s2. When the power of the residual echo is smaller than a predetermined level or more, it is determined that the impulse response of the adaptive filter 61 matches the actual echo transmission system more than the impulse response of the fixed filter 62, or the input transmission signal i The level of the echo suppression amount (ERLE; Echo Return Loss Enhancement) for the error signal e1 due to the first pseudo echo signal s1 is the level of the echo suppression amount ERLE for the error signal e2 due to the second pseudo echo signal s2 of the input transmission signal i. If the level is larger than a predetermined level, the impulse response of the adaptive filter 61 is smaller than that of the fixed filter 62. It has been made by determining to be matched to the actual echo train than the impulse response.
[0007]
The double talk detection unit 63 detects double talk, that is, a state in which the near-end user is speaking, based on the level difference between the line reception signal o and the line transmission signal e2, and detects double talk. The operation of transferring the coefficients of the adaptive filter 61 to the fixed filter 62 during the period is stopped. This is because, when the near-end user is speaking, since the input transmission signal i contains many components other than the echo component due to the user's speech, the adaptive filter 61 calculates the coefficients appropriately. Is no longer possible.
[0008]
Now, there is a form called a hands-free call as a form of telephone conversation of a user using a telephone, which is a typical example of voice communication.
In the hands-free call, a telephone call is performed by using a microphone and a speaker outside the telephone to input a transmission voice and output a reception voice. Here, a device that includes a speaker that radiates a reception voice into the vehicle interior space and a microphone that picks up the voice of the vehicle interior space, and adds the above-described hands-free communication function to a mobile phone is called a hands-free communication device or the like. I have. Such a hands-free communication device is mainly installed in a car, and is used for enabling a driving user to make a hands-free call using a mobile phone.
[0009]
Here, in such a hands-free call, since the received voice output from the speaker goes around the microphone (echo) at a relatively large level, it is extremely important to appropriately cancel the echo by an echo canceller. It becomes.
[0010]
[Problems to be solved by the invention]
Now, for example, when the echo canceller shown in FIG. 6 is applied to the above-described hands-free communication in a car, the following problem occurs.
That is, in the automobile, the traveling sound and other car noises are mixed with the input transmission signal i at a relatively large level together with the echo. The influence of the car noise appears in the error signal e1 between the input transmission signal i and the first pseudo echo signal s1 and the error signal e2 between the input transmission signal i and the second pseudo echo signal s2. According to the conventional echo canceler technology, even when the adaptive filter 61 generates a pseudo echo signal more suitable for echo cancellation, its coefficient is not immediately transferred to the fixed filter 62, and as a result, It may take a relatively long time before the echo is sufficiently canceled. Further, in an environment where car noise is present, the coefficients of the adaptive filter 61 are sequentially updated, and the echo components included in the error signal e1 between the input transmission signal i and the first pseudo echo signal s1 can be heard as audio sound. On the other hand, depending on the operation of the fixed filter 62 in which the coefficient is updated intermittently, the echo component included in the error signal e2 between the input transmission signal i and the second pseudo echo signal s2 is perceived as audio sound. In some situations, it becomes audible.
[0011]
In other words, according to the conventional echo canceler technology, the pseudo echo signal generated by the fixed filter 62 necessarily has a higher line quality than the pseudo echo signal generated by the adaptive filter 61 in terms of auditory quality. This is not always suitable for echo cancellation, and for this reason, there is a case where it is not possible to perform high-quality echo cancellation on hearing.
[0012]
Therefore, an object of the present invention is to provide an echo canceller capable of performing higher quality echo cancellation in aural sense.
[0013]
[Means for Solving the Problems]
To achieve the above object, the present invention provides an echo canceller that cancels an echo of a first input signal from a second input signal, an adaptive filter that generates and outputs a first pseudo echo signal from the first input signal, A first adder for subtracting the first pseudo echo signal from the second input signal to generate a first transmission signal, a fixed filter for generating and outputting a second pseudo echo signal from the first input signal, and a second input A second adder for subtracting the second pseudo echo signal from the signal to generate a second transmission signal, and selecting means for selecting and outputting one of the first transmission signal and the second transmission signal. It is composed. Here, the adaptive filter updates a coefficient for determining a first pseudo echo signal generated from a first input signal according to the first transmission signal, and the fixed filter is transferred from the adaptive filter. A second pseudo echo signal is generated from the first input signal according to the coefficient.
[0014]
According to such an echo canceller, it is possible to appropriately switch and output the second transmission signal echo-cancelled by the fixed filter and the first transmission signal echo-cancelled by the adaptive filter.
Therefore, even if it is not possible to perform high-accuracy echo cancellation by the above-described double filter algorithm, if the high-accuracy echo cancellation can be performed by the adaptive filter, the adaptive filter performs echo cancellation. By outputting the first transmission signal obtained as described above, it is possible to perform acoustically high-quality echo cancellation.
[0015]
More specifically, according to such an echo canceller, when the convergence state of the adaptive filter is stable, or when the time variation of the echo suppression amount in the first transmission signal by the adaptive filter is smaller than a predetermined level. The time variation of the coefficient of the adaptive filter is smaller than a predetermined level; the amount of echo suppression in the first transmission signal by the adaptive filter is increasing; or the environmental noise sound included in the second input signal. It can be estimated that when the size of is larger than a predetermined level, the use of the adaptive filter can perform more acoustically high-quality echo cancellation. Therefore, in these cases, it can be expected that a higher auditory quality echo cancellation can be realized by causing the selecting means to select the first transmission signal.
[0016]
Here, such an echo canceller can be applied to, for example, a call voice processing device that processes a call voice of a voice communication terminal. That is, the call voice processing device is composed of, for example, the above-described echo canceller, a speaker for outputting the received voice received by the voice communication terminal, and a microphone for inputting the voice. The first input signal and the voice input from the microphone may be operated as the second input signal, and the output of the selection means may be the transmission voice transmitted by the voice communication terminal.
[0017]
Further, when the present echo canceller is applied to such a call voice processing device, a line state detecting unit for detecting a connection state of a call line used for a call by the voice communication terminal is provided, and the control unit includes: When the line state detecting means detects the connection of the communication line, the coefficient of the adaptive filter may be transferred to the fixed filter. From the start of transmission / reception to the line connection, the near-end user does not speak and enters a non-double talk state, so that it can be expected that an appropriate coefficient will be set in the adaptive filter at the time of line connection. Thus, by doing so, it becomes possible to set an appropriate coefficient to the fixed filter at the time of line connection and thus at the start of a call.
[0018]
Alternatively, a double talk detecting means for detecting a speech state of the user of the telephone call processing apparatus is provided, and the control unit may be configured to, when the double talk detection means detects the speech state of the user, perform the first When the transmission signal and the second transmission signal are different from each other by a predetermined level or more, the coefficient of the fixed filter may be transferred to the adaptive filter. In this way, when the coefficient of the adaptive filter greatly deviates from an appropriate value due to double talk, the coefficient of the adaptive filter can reach the appropriate coefficient more quickly. .
[0019]
According to another aspect of the present invention, there is provided an echo canceller for canceling an echo of a first input signal from a second input signal, wherein the first pseudo echo signal is generated and output from the first input signal. A filter, a first adder for subtracting the first pseudo echo signal from the second input signal to generate a first transmission signal, and a fixed filter for generating and outputting a second pseudo echo signal from the first input signal; A second adder for subtracting a second pseudo echo signal from a second input signal to generate a second transmission signal, and transferring coefficients of the adaptive filter to the fixed filter in accordance with a time variation of coefficients of the adaptive filter. The adaptive filter updates a coefficient for generating a first pseudo echo signal from a first input signal in accordance with the first transmission signal, and the fixed filter performs adaptive control. From the filter Also echo canceller to the first input signal in accordance with transmission coefficient and generates the second pseudo echo signal provides.
[0020]
According to such an echo canceller, whether the coefficient of the current adaptive filter is more suitable for echo cancellation than the coefficient of the fixed filter, more appropriately taking into account the stability of the adaptive filter. It is possible to determine whether or not to transfer the coefficient.
[0021]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an echo canceller according to the present invention will be described by taking an example of application to an in-vehicle hands-free communication device.
FIG. 1 shows a configuration of a hands-free communication device according to the present embodiment.
As shown in the figure, the hands-free communication device includes a voice processing device 1, a speaker 2, a microphone 3, and a telephone interface unit 4. Further, a mobile telephone 5 is connected to the telephone interface unit 4.
In such a configuration, the telephone interface unit 4 is detachably connected to the mobile phone 5, outputs a reception signal R input from the mobile phone 5 to the voice processing device 1, and transmits a transmission transmission signal input from the voice processing device 1. The signal T is output to the mobile phone 5. In addition, the telephone interface unit 4 is provided with an input device such as a switch, relays an on-hook / off-hook command to the mobile phone 5 according to a user operation of the input device, and causes the mobile phone 5 to perform an on-hook / off-hook operation. For example, processing for notifying the voice processing device 1 of the outgoing / incoming call state and line connection state of the mobile phone 5 is performed.
[0022]
Next, the audio processing device 1 includes an adaptive filter 11, a fixed filter 12, a double-talk detector 13, a selector 14, a switching controller 15, a first adder 16, a second adder 17, and an amplifier 18.
In such a configuration, the amplifier unit 18 drives the speaker 2 with the output signal o obtained by amplifying the reception signal R input from the telephone interface unit 4, and outputs sound to the vehicle interior space.
On the other hand, the microphone 3 picks up the sound in the vehicle interior space and sends it to the first adder 16, the second adder 17, and the switching control unit 15 as a transmission signal i. The first adder 16 subtracts the first pseudo echo signal s1 calculated by the adaptive filter 11 from the input transmission signal i, and outputs the result to the selector 14 as a first transmission signal e1. Then, the second adder 17 subtracts the second pseudo echo signal s2 calculated by the fixed filter 12 from the input transmission signal i from the transmission signal, and outputs the result to the selector 14 as the second transmission signal e2. Output. The selector 14 selects one of the first transmission signal e1 input from the first adder 16 and the second transmission signal e2 input from the second adder 17, and transmits the transmission transmission signal to the telephone interface unit 4. Output as a signal T.
[0023]
The adaptive filter 11 includes a coefficient update unit 111 and a filter unit 112. Then, the coefficient update unit 111 uses a learning identification method such as an LMS (Least Mean Square) algorithm or an NLMS (Normalized Least Mean Square) algorithm to minimize the filter of the adaptive filter 11 so that the power of the first adder output e1 is minimized. The processing of calculating the coefficient and setting the coefficient in the filter unit 112 is repeated. The filter unit 112 is, for example, an FIR filter having an impulse response determined by the coefficient set by the coefficient update unit 111, and generates the first pseudo echo signal s1 from the reception signal R.
[0024]
Next, the fixed filter 12 is a filter to which the coefficients of the adaptive filter 11 are transferred, has the same configuration as the filter unit 112, and generates the second pseudo echo signal s2 from the reception signal R according to the transferred coefficients. . Further, the double talk detecting unit 13 detects a double talk state according to a level difference between the reception signal R and the second transmission signal e2, and notifies the switching control unit 15 of the detected double talk state.
[0025]
Here, in such an audio processing unit, the switching control unit 15 controls the transfer of coefficients from the adaptive filter 11 to the fixed filter 12 and the selection operation of the selector 14.
Hereinafter, the control operation of the switching control unit 15 will be described.
First, FIG. 2 shows a procedure of a switching control process of the selector 14 performed by the switching control unit 15.
As shown in the figure, in this processing, the coefficient of the adaptive filter 11 is obtained (step 202), and the coefficient error CE of the adaptive filter 11 is set to ω (i), the i-th tap coefficient of the current filter unit 112, ωn ( i) as the i-th tap coefficient of the filter unit 112 a predetermined time before the present time, a process of calculating by the expression 1 or the expression 2 approximating the expression 1 (step 204), and the echo suppression amount ERLE (Echo A process of calculating Return Loss Enhancement (unit db) using the transmission signal i and the first transmission signal e1 using e3 as e1 = e (step 206), and the amount of suppression ERLE by the fixed filter 12 is calculated using the transmission signal Using i and the second transmission signal e2, the process of calculating by equation 3 with e2 = e (step 208) is repeated.
[0026]
(Equation 1)

Then, a car noise component is extracted from the transmission signal i by using an appropriate filter, and it is checked whether or not the level of the car noise is higher than a predetermined level (step 210). Is controlled (step 214). This is because, when the level of the car noise is higher than a certain level, the first transmission signal e1 generated by the adaptive filter 11 is selected to be higher than the second transmission signal e2 by the fixed filter 12 as described above. This is because there is a high possibility that quality echo cancellation can be performed. Here, the predetermined level of the car noise serving as the threshold for switching the selection to the first transmission signal e1 is, for example, experimentally obtained and set in advance.
[0027]
On the other hand, if the level of the Kerr noise is not equal to or higher than the predetermined level, it is determined whether the convergence state of the adaptive filter 11 is stable at that time. It is determined whether the time variation of the echo suppression amount ERLE is smaller than a predetermined level (step 212). If the convergence state of the adaptive filter 11 is stable, the selector 14 outputs the first transmission signal e1. Control is performed (step 214). This is because if the convergence state of the adaptive filter 11 is stable, it can be considered that the acoustically high-quality echo cancellation is realized by the adaptive filter 11 in the first transmission signal e1.
[0028]
Here, the coefficient error CE has a characteristic that increases as the level of the first transmission signal e1 increases, and has a constant correlation with the echo suppression amount ERLE by the adaptive filter 11, for example, as shown in FIG.
Next, if the level of the Kerr noise is not higher than the predetermined level and the convergence state of the adaptive filter 11 is not stable, it is determined whether or not the echo suppression amount ERLE by the adaptive filter 11 tends to increase with time. Investigation (step 216), if there is an increasing trend, the selector 14 is controlled to output the first transmission signal e1 (step 214); otherwise, the second transmission signal e2 is output. (Step 218). This is because, when the echo suppression amount ERLE by the adaptive filter 11 increases, the first transmission signal e1 is larger than the second transmission signal e2 by the fixed filter 12 to which the past coefficient of the adaptive filter 11 is applied. This is because, in the present or very near future, more acoustically high quality echo cancellation may be realized.
[0029]
Next, FIG. 4 shows a procedure of a coefficient transfer control process performed by the switching control unit 15.
As shown in the figure, in this processing, it is checked whether or not the line connection state has changed from the line connection state notified from the telephone interface unit 4 to a line connected state (step 402), and the state changes to the line connected state. If so, the coefficients of the adaptive filter 11 are transferred to the fixed filter 12 (step 410). This is because the near-end user does not speak during the period from the start of the transmission / reception to the line connection and is in the non-double-talk state. Therefore, an appropriate coefficient will be set in the adaptive filter 11 at the time of the line connection. It is because that can be expected.
[0030]
In this process, it is checked whether the double talk detecting unit 13 is detecting the double talk state (step 404). If the double talk state is not being detected, the ERLE of the adaptive filter 11 is higher than the ERLE of the fixed filter 12. If good (step 406), the coefficients of the adaptive filter 11 are transferred to the fixed filter 12 (step 410).
[0031]
However, in the above step 406, it may be determined whether to transfer the coefficients of the adaptive filter 11 to the fixed filter 12 in consideration of the coefficient error CE together with the ERLE. That is, for example, the coefficients of the adaptive filter 11 are transferred to the fixed filter 12 only when the ERLE of the adaptive filter 11 is better than the ERLE of the fixed filter 12 and the coefficient error CE is smaller than a predetermined level. When the value obtained by multiplying the ERLE of the fixed filter 11 by a certain degree to the ERLE of the fixed filter 12 and the degree of the coefficient error CE is a predetermined level or more, the coefficients of the adaptive filter 11 are transferred to the fixed filter 12. It may be. In this way, by considering the coefficient error CE together with the ERLE, the coefficient of the truly current adaptive filter 11 is more appropriately set in consideration of the stability of the adaptive filter 11 than the coefficient of the current fixed filter 12. It is possible to determine whether or not the signal is suitable for echo cancellation. The transfer control of the coefficient in consideration of the coefficient error CE can be similarly applied to the transfer control of the coefficient from the adaptive filter 11 to the fixed filter 12 in the conventional echo canceller of the double filter algorithm shown in FIG. is there.
[0032]
On the other hand, when the double talk detecting unit 13 is detecting the double talk state, it is determined whether or not the levels of the first transmission signal e1 and the second transmission signal e2 are different from each other by a predetermined level or more (for example, twice or more). (Step 412) If different, the coefficients of the fixed filter 12 are transferred to the adaptive filter 11 (Step 414), and set in the filter unit 112. Also, when the ERLE of the fixed filter 12 is several times (for example, 4 to 8 times) better than the ERLE of the adaptive filter 11 (step 408), the coefficient of the fixed filter 12 is transferred to the adaptive filter 11 (step 408). Step 414), set in the filter unit 112.
[0033]
Further, even when the double talk detecting unit 13 is not detecting the double talk state, if the ERLE of the fixed filter 12 is several times (for example, 4 to 8 times) better than that of the adaptive filter 11 ( Step 408), the coefficients of the fixed filter 12 are transferred to the adaptive filter 11 (Step 414), and set in the filter unit 112. In such a case, the coefficient of the adaptive filter 11 is relatively large and erroneous, and it is better to return the coefficient of the fixed filter 12 to the adaptive filter 11 and perform the coefficient update operation from the coefficient to obtain an appropriate coefficient. Is considered to be able to be more promptly reached.
[0034]
The embodiments of the present invention have been described above.
By the way, in the above, the present embodiment has been described taking an example of application to an in-vehicle hands-free communication device, but the echo canceller according to the present embodiment can be applied to any voice communication device. Further, the echo canceller according to the present embodiment can be applied to echo cancellation in any audio processing device 1 other than the audio communication device. For example, cancellation of the output sound from the speaker 2 of the in-vehicle audio device or the navigation device 52 to the voice input from the microphone 3 to the voice recognition device 51 as shown in FIG. The present invention can be similarly applied to, for example, wrapping of karaoke performance music output from the speaker 2 to audio input from the speaker. However, when the present invention is applied to the audio processing device 1 used in an environment other than the vehicle, in step 210 in FIG. 2, noise generated in the environment is used for switching the output of the selector 14 instead of the car noise.
[0035]
【The invention's effect】
As described above, according to the present invention, it is possible to provide an echo canceller capable of performing higher quality acoustically canceling echo.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of a hands-free communication device according to an embodiment of the present invention.
FIG. 2 is a flowchart illustrating a procedure of output switching control of a selector according to the embodiment of the present invention.
FIG. 3 is a diagram showing a correlation between a coefficient error and ERLE used in the embodiment of the present invention.
FIG. 4 is a flowchart illustrating a procedure of coefficient transfer control according to the embodiment of the present invention.
FIG. 5 is a diagram showing another application of the echo canceller according to the embodiment of the present invention.
FIG. 6 is a block diagram showing a configuration of a conventional echo canceller.
[Explanation of symbols]
1: audio processing device, 2: speaker, 3: microphone, 4: telephone interface device, 5: mobile phone, 11, 61: adaptive filter, 12, 62: fixed filter, 13, 63: double talk detection unit, 14: Selector, 15: switch control unit, 16, 64: first adder, 17, 65: second adder, 18: amplifier unit, 66: double filter control unit, 112: filter unit.

Claims (11)

An echo canceller for canceling an echo of a first input signal from a second input signal,
An adaptive filter that generates and outputs a first pseudo echo signal from the first input signal, a first adder that subtracts the first pseudo echo signal from the second input signal to generate a first transmission signal, A fixed filter that generates and outputs a second pseudo echo signal from an input signal, a second adder that subtracts a second pseudo echo signal from a second input signal to generate a second transmission signal, and the first transmission Selecting means for selecting and outputting one of the signal and the second transmission signal,
The adaptive filter updates a coefficient for generating a first pseudo echo signal from a first input signal according to the first transmission signal, and the fixed filter performs a first input according to a coefficient transferred from the adaptive filter. An echo canceller for generating a second pseudo echo signal from a signal. 2. The echo canceller according to claim 1, wherein said selecting means includes a control unit for selecting said first transmission signal when the convergence state of said adaptive filter is stable. The echo canceller according to claim 2, wherein
The echo canceller, wherein the control unit causes the selection unit to select the first transmission signal when the time variation of the echo suppression amount in the first transmission signal by the adaptive filter is smaller than a predetermined level. . The echo canceller according to claim 2, wherein
The echo canceller, wherein the control unit causes the selecting unit to select the first transmission signal when the time variation of the coefficient of the adaptive filter is smaller than a predetermined level. The echo canceller according to claim 1, wherein
An echo canceller, characterized in that the selecting means has a control unit for selecting the first transmission signal when the amount of echo suppression in the first transmission signal by the adaptive filter is increasing. The echo canceller according to claim 1, wherein
An echo canceller comprising: a control unit that causes the selection unit to select the first transmission signal when the magnitude of the environmental noise sound included in the second input signal is higher than a predetermined level. A call voice processing device for processing a call voice of a voice communication terminal,
An echo canceller according to claim 2, 3, 4, 5, or 6,
A speaker for outputting a received voice received by the voice communication terminal;
A microphone for inputting voice,
The echo canceller operates the received voice as the first input signal and the voice input from the microphone as the second input signal, and uses the output of the selecting means as the transmitted voice transmitted by the voice communication terminal. A telephone call processing device characterized by being performed. The telephone call processing device according to claim 7,
Having a line state detecting means for detecting a connection state of a call line used for a call by the voice communication terminal,
The communication voice processing apparatus according to claim 1, wherein the control unit causes the coefficient of the adaptive filter to be transferred to the fixed filter when the line state detection unit detects the connection of the communication line. The telephone call processing device according to claim 7 or 8,
Having double talk detection means for detecting the state of the user of the call voice processing device during speech,
The control unit is configured to, when the double talk detecting unit detects the state of the user being uttered, when the first transmitted signal and the second transmitted signal are different from each other by a predetermined level or more, the coefficient of the adaptive filter is Is updated to the coefficient of the fixed filter. The telephone call processing device according to claim 7, 8, or 9,
A telephone call processing apparatus comprising a mobile telephone interface for accommodating a mobile telephone as the voice communication terminal. An echo canceller for canceling an echo of a first input signal from a second input signal,
An adaptive filter that generates and outputs a first pseudo echo signal from a first input signal; a first adder that subtracts the first pseudo echo signal from a second input signal to generate a first transmission signal; A fixed filter that generates and outputs a second pseudo echo signal from the signal, a second adder that subtracts the second pseudo echo signal from the second input signal to generate a second transmission signal, and a time of a coefficient of the adaptive filter. A control unit that controls the transfer of the coefficients of the adaptive filter to the fixed filter in accordance with the variation,
The adaptive filter updates a coefficient for generating a first pseudo echo signal from a first input signal according to the first transmission signal, and the fixed filter performs a first input according to a coefficient transferred from the adaptive filter. An echo canceller for generating a second pseudo echo signal from a signal.

Images