JP3616341B2 - Multi-channel echo cancellation method, apparatus thereof, program thereof, and recording medium - Google Patents

Multi-channel echo cancellation method, apparatus thereof, program thereof, and recording medium Download PDF

Info

Publication number
JP3616341B2
JP3616341B2 JP2001051387A JP2001051387A JP3616341B2 JP 3616341 B2 JP3616341 B2 JP 3616341B2 JP 2001051387 A JP2001051387 A JP 2001051387A JP 2001051387 A JP2001051387 A JP 2001051387A JP 3616341 B2 JP3616341 B2 JP 3616341B2
Authority
JP
Japan
Prior art keywords
channel
signal
echo
pseudo
impulse response
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
JP2001051387A
Other languages
Japanese (ja)
Other versions
JP2002261659A (en
Inventor
朗 中川
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nippon Telegraph and Telephone Corp
Original Assignee
Nippon Telegraph and Telephone Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nippon Telegraph and Telephone Corp filed Critical Nippon Telegraph and Telephone Corp
Priority to JP2001051387A priority Critical patent/JP3616341B2/en
Publication of JP2002261659A publication Critical patent/JP2002261659A/en
Application granted granted Critical
Publication of JP3616341B2 publication Critical patent/JP3616341B2/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Images

Landscapes

  • Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
  • Telephone Function (AREA)

Description

【0001】
【発明の属する技術分野】
この発明は、多チャネル再生系を有する例えば通信会議システムにおいて、ハウリングの原因および聴覚上の障害となる室内反響信号を消去する多チャネル反響消去方法、その装置、そのプログラム及びその記録媒体に関するものである。
【0002】
【従来の技術】
同時通話性能に優れ反響感の少ない拡声通話システムを提供するために、反響消去装置がある。
まず、1チャネル用の反響消去装置について、その反響消去方法および装置構成を図4を参照して説明する。拡声通話において、受話端子11に相手の発話等で得られる受話信号は、そのままスピーカ12から再生される場合と、スピーカ12へ送る前に、受話信号の振幅やパワー等の大きさに応じて自動的に利得を調節するなど、受話信号に何らかの加工が施された後に、スピーカ12から再生される場合とがある。このため、この明細書で受話信号x(k)とは、相手からの受話信号そのものとは限らず、受話信号に対し加工が施された場合は、その加工された後の受話信号を指すものとする。kは離散時間を表す。反響消去装置14は、受話信号x(k)がスピーカ12から反響路15を経て、マイクロホン16に集音されて得られる反響信号y(k)を消去する。ここで、反響信号y(k)は、時刻kにおける反響路15のインパルス応答をh11(k,n)として、

Figure 0003616341
Σはn=0からL−1まで
のような畳み込み演算で得られるものとモデル化できる。Lはタップ数で、反響路15の残響時間に対応させて、あらかじめ設定しておく定数である。まず、受話信号蓄積・ベクトル生成部17において、受話信号x(k)をL−1時刻過去のものまで蓄積しておく。蓄積された信号は、受話信号ベクトル (k)、すなわち
Figure 0003616341
として出力される。但し、*はベクトルの転置を表す。疑似反響信号生成部18では、式(2)の受話信号ベクトル (k)と、反響路推定部19から得られる疑似反響ベクトル11(k)との内積演算
Figure 0003616341
を行ない、その結果として、疑似反響信号y^(k)を生成する。この内積演算は、式(1)のような畳み込み演算と等価である。反響路推定部19では、疑似反響信号生成部18で用いる疑似反響路ベクトル11(k)を生成する。この反響路推定に用いる最も一般的なアルゴリズムは、NLMSアルゴリズム(学習同定法)である。NLMSアルゴリズムでは、時刻kにおける受話信号ベクトル (k)と、残留反響信号e(k)、すなわち差回路21でマイクロホン16の出力y(k)から疑似反響信号y^(k)を差し引いた信号
Figure 0003616341
とから、時刻k+1において用いる疑似反響路ベクトル11(k+1)を次式のように求める。
Figure 0003616341
但し、μはステップサイズパラメータと呼ばれ0<μ<2の範囲で適応動作の調整に用いる。以上のような処理を繰り返すことにより、反響路推定部19では、次第に疑似反響路ベクトル11(k)を、真の反響路15のインパルス応答h11(k,n)の時系列を各要素として持つ反響路ベクトル 11(k)、すなわち
Figure 0003616341
と一致させることが可能となり、その結果式(4)の残留反響信号e (k)を小さくすることができる。
【0003】
一般にN(2)チャネルの再生系とM(1)チャネルの集音系とで構成される通信会議システムの場合の反響の消去は、図5に示すような構成により行われる。すなわち再生側の全Nチャネルと集音側の各1チャネルとの間にN入力1出力時系列信号を処理するNチャネル反響消去装置22,22,…,22をそれぞれ接続した反響消去システム23として実現される。この場合システム全体でN×M個の反響路15nm(1N,1M)が存在する。このシステムの構成単位である再生側の全Nチャネルと集音側の各1チャネルとの間に接続されるNチャネル反響消去装置22,22,…,22については、図4に示した反響消去装置14の構成を拡張して、図6に示すように構成される。これは例えば電子情報通信学会論文誌 ’86/10 Vol.J69−A No.10「多チャネル適応ディジタルフィルタ」に詳しく述べられている。ここで、集音側が第m集音チャネル(1M)に接続されているNチャネル反響消去装置22を考える。第mチャネルのマイクロホン16で集音される反響信号は、各再生チャネルの受話信号がそれぞれの反響路151m〜15Nmを経て集音側で全て加算されることにより得られるために、反響路推定をどの再生チャネルについても、統一的に同じ1つの残留反響信号e(k)のみを評価して行なうための工夫が必要となる。まず、各再生チャネルの受話信号について、受話信号蓄積・ベクトル生成部(17,17,…,17)により、受話信号ベクトル
Figure 0003616341
を生成する。但し、L,L,…,Lはタップ数で、各反響路151m,152m,…,15Nmの残響時間に対応させて、あらかじめ設定する定数である。これらのベクトルをベクトル結合部24によって、
Figure 0003616341
と結合する。また、反響路推定部19においても、各再生チャネルと第m集音チャネルとの間のN個の反響路を模疑するための、各疑似反響路ベクトル1m(k),2m(k),…,Nm(k) を結合して
Figure 0003616341
として扱う。疑似反響路結合ベクトル(k) の更新は、NLMSアルゴリズムを用いた場合、
Figure 0003616341
のように行なわれる。疑似反響信号生成部18では、内積演算
Figure 0003616341
により、第m集音チャネルで集音された反響信号y(k)に対する疑似反響信号y^(k)を生成する。このように、各チャネル毎のベクトルを結合して1つのベクトルとして扱うことにより、基本的な処理の流れは、図4に示した1チャネル反響消去装置と同様となる。
【0004】
【発明が解決しようとする課題】
従来の多チャネルエコー消去方法はその疑似反響路が真の反響路にほぼ収束するのが遅い。この点を改善すべく研究した結果、各チャネルの受話信号の再生信号ごとの反響路についての疑似反響路において、その反響路の逐次推定における更新ステップサイズμが同一値である点に問題があると考えるに至った。
説明を簡単にするため、図7に示すように受話チャネルが2つ、集音チャネルが1つの場合を例とする。各受話信号x(k),x(k)はそれぞれチャネル疑似反響路18,18へ印加され、チャネル疑似反響路18,18よりの各チャネル疑似反響信号y^(k),y^(k)は合成回路26で合成されて総合疑似反響信号y^(k)とされて差回路21へ供給され、マイクロホン16よりの集音信号y(k)からy^(k)が引算され、残留反響信号e(k)が求められる。残留反響信号e(k)と、受話信号x(k),x(k)がチャネル反響路推定部19,19へ供給され、チャネル反響路推定部19,19はそれぞれ、受話信号x(k),x(k)の各再生信号がマイクロホン16に至る反響路15,15のインパルス応答 (k), (k)を逐次推定して、これら推定したインパルス応答(k),(k)をチャネル疑似反響路18,18にそれぞれ設定する。
【0005】
チャネル反響路推定部19,19における各推定チャネルインパルス応答(k),(k)は式(12)から以下のようになる。式(12)は受話チャネルが2つであるから式(14)となる。
【0006】
【数1】
Figure 0003616341
【0007】
反響路151 ,152 の残響時間により決るタップ数L1 とL2 を等しくLとした。
このようにXT(k)X(k)はスカラーであるから、式(14)は行列の各行ごとに次のように書ける。
Figure 0003616341
つまりチャネル反響路推定部191 ,192 はそれぞれ式(15),(16)によりチャネル反響路151 ,152 のインパルス応答h1(k),h2(k)を逐次推定することになる。X1(k)/(XT(k)X(k)),X2 (k)/(XT(k)X(k))はそれぞれ全チャネルの受話信号のパワーで正規化された受話信号(ベクトル)と言える。XT(k)X(k)はパワー計算部27で計算されて各推定部191 ,192 へ供給される。また図中には図中の受話信号蓄積ベクトル生成部は必要とするが省略してある。式(15),(16)より、その更新ステップサイズは共にμと等しいものとなっている。
【0008】
しかし、例えば図7に示すように、スピーカ12,12とマイクロホン16,16を互いに対向させて、方形の各コーナに位置させた場合、マイクロホン16に対し正面のスピーカ12からの反響路15のインパルス応答h(k)の方が、対角位置のスピーカ12からの反響路15のインパルス応答 (k)よりエネルギーが大きい。このことは、スピーカとマイクロホンの方形コーナの配置に限らず、他の配置においても反響路によりインパルス応答のエネルギーに差があり、かつ3チャネル以上の多チャネルエコー消去方法についても同様のことが言える。
【0009】
そこでこの発明の目的はこのインパルス応答のエネルギーの差を考慮して従来よりも収束が速い多チャネルエコーキャンセル方法、その装置、そのプログラム及びその記録媒体を提供することにある。
【0010】
【課題を解決するための手段】
この発明によれば各チャネルの受話信号をチャネル疑似反響路に印加してチャネル疑似反響信号をそれぞれ生成し、これらチャネル疑似反響信号を加算して総合疑似反響信号を生成し、この総合疑似反響信号を集音信号から差し引いて残留反響信号を生成し、特に各チャネルの受話信号の再生信号と集音信号との各インパルス応答と対応する信号の大きさを入力して、各チャネルのインパルス応答の級数の係数の二乗和の相対比と各チャネルの更新ステップサイズの相対比が等しくなるように各チャネルの更新ステップサイズを生成し、各チャネルごとにその更新ステップサイズと、各受話信号と、上記残留反響信号とからそのチャネルの反響路のインパルス応答を逐次推定し、その推定したインパルス応答によりそのチャネルの疑似反響路を生成する。
【0011】
【発明の実施の形態】
図1にこの発明の実施形態を示す。説明を簡単にするため、受話2チャネル、集音1チャネルを例とし、図7と対応する部分に同一参照符号を付けてある。この実施形態では、各チャネルの反響路15,15のインパルス応答 (k), (k)と対応する信号をステップサイズ生成部31に入力して、各チャネルごとのステップサイズμ,μをそのインパルス応答に対応する信号の大きさに応じて決定する。インパルス応答 (k), (k)と対応する信号としては受話信号x(k),x(k)の各再生信号と集音信号y(k)との間の各インパルス応答、つまり反響路15,15の各インパルス応答 (k), (k)それ自体、あるいはスピーカ12,12、マイクロホン16の配置に応じて経験的に推測される大略の値、またはチャネル疑似反響路18,18に設定される推定インパルス応答などが用いられる。インパルス応答 (k), (k)を用いる場合は、予め、各1つのスピーカのみからインパルス音を放射して、マイクロホン16の出力信号を測定して求める。
【0012】
更新ステップサイズμ,μは各チャネルのインパルス応答 (k), (k)と対応する信号(以下 (k), (k)で代表する)の大きさ、つまり振幅又はパワーに応じ、大きければ大きな値とする。特に各チャネルのインパルス応答 (k), (k)のパワーの、これらパワーの和に対する比に応じて更新ステップサイズμ,μを決定する。この場合、これら各チャネルの更新ステップサイズμ とμ の和が2となるようにする。例えば次のようにして求める。いま (k), (k)がそれぞれ1要素のみとして、これらを平面上に書くと図2に示すように、式(14)の更新ベクトル^(k)は、 (k)と (k)のパワーが等しい場合は、横軸に対し45度のベクトル41上にあることになる。ベクトル (k), (k)の各長さをa=√( (k) (k)),b=√( (k) (k))とすると、更新ベクトル^(k)は図2中のベクトル42上のように設定することで、その疑似反響路が真の反響路にほぼ収束するのを速くすることができる。この更新ベクトル^(k)に応じてステップサイズμ,μを、√(a)と√(b)の比と等しくする、つまり
μ:μ=√(a):√(b
とすればよい。従って下記の連立方程式が得られる。
【0013】
μ1 √(b2 )=μ2 √(a2
μ1 2+μ2 2=2
この連立方程式を解くと以下のようになる。
μ1√(2a 2 )/(√(a 2 +b 2 ))
μ2√(2b 2 )/(√(a 2 +b 2 ))
つまり
μ1√(21 T(k)h1(k))/(√(h1 T(k)h1(k))+√(h2 T(k)h2 (k))) (17)
μ2√(22 T(k)h2(k))/(√(h1 T(k)h1(k))+√(h2 T(k)h2 (k))) (18)
をステップサイズ生成部31で計算すればよい。
【0014】
従ってインパルス応答 (k), (k)がそれぞれ2乗和回路32,32に入力されて、各ベクトルの要素の2乗和
(k) (k)=Σ i=0 L−1 (k,i)
(k) (k)=Σ i=0 L−1 (k,i)
がそれぞれ計算され、これら2乗和は開平回路33,33で平方根がとられ、a=√( (k) (k),b=√( (k) (k))が求められる。これらaとbが加算回路34で加算され、式(17),(18)の分母が求められる。乗除算回路35で加算回路34の加算値でaを除算した値に2を乗算して式(17)の計算結果μが求められる。同様に乗除算回路35で加算回路34の加算値でbを除算した値に2を乗算して式(18)の計算結果μが求められる。
【0015】
これら更新ステップサイズμ,μがそれぞれ対応するチャネル反響推定部19,19へ供給される。チャネル反響路推定部19,19には図6に示した場合と同様にそれぞれ受話信号x(k),x(k)が供給されると共にパワー計算部27よりの所定時間の全チャネルの受話信号パワーの和と残留反響信号e(k)が共通に供給される。チャネル反響路推定部19,19ではそれぞれ以下の式によりチャネル反響路のインパルス応答の推定(k+1),(k+1)が逐次行われる。
【0016】
Figure 0003616341
これら推定されたインパルス応答(k+1),(k+1)がそれぞれチャネル疑似反響路18,18に設定される。その他は図6に示した装置と同様な処理を行う。
【0017】
このように更新ステップサイズμ,μが反響路15,15のインパルス応答 (k), (k)のエネルギーの相対比に応じて、大きいエネルギーのチャネルに対応する更新ステップサイズが大きくされ、それだけ全体としての疑似反響路18,18の収束が速くなる。
ステップサイズ生成部31で各時刻kごとにμ(k),μ(k)を計算することにより、反響路15,15が変動して (k), (k)が変動してもこれに追従して、適切なμ(k),μ(k)が設定されることになる。また最初にステップサイズ生成部31に、インパルス応答 (k), (k)の経験的推定値を入力しても、その後、チャネル疑似反響路18,18の各インパルス応答(k),(k)をステップサイズ生成部31に入力することにより、チャネル疑似反響路18,18が反響路15,15に収束するに従って、更新ステップサイズμ,μが適切な値μ(k),μ(k)に更新される。あるいはμ,μの初期値はμ=μとし従来において使用していた適当な値を設定し、その後は、(k),(k)をステップサイズ生成部31に入力してμ,μを更新してもよい。
【0018】
この発明は受話チャネルが3チャネル以上の場合にも適用できる。例えばNチャネルの場合、ステップサイズ生成部31で各チャネルn(n=1,2,・・・,N)に対し、次の計算をすればよい。
μn√N(hn T(k)hn(k))/√(h1 T(k)h1(k)+(h2 T(k)h2(k)・・・+hN T(k)hN(k))
この式から理解されるように、各チャネルのインパルス応答の大きさ(振幅又はパワー)の、チャネル間の相対的大きさ比により、インパルス応答の大きさが大きいチャネルの疑似反響路の収束速度が大きくなるようにステップサイズを求めればよい。また上記実施形態ではチャネル反響路推定部において、全チャネルの受話信号のパワーの和を求めたものを用いたが、必ずしもその必要はない。例えば電子情報通信学会論文誌’86/10 Vol.J69−A No.10「多チャネル適応ディジタルフィルタ」で提案しているように、チャネルごとにその受話信号をその電力で正規化したものを用いてもよい。
【0019】
またチャネル反響路推定のアルゴリズムとしては学習同定法に限らず、射影法その他のアルゴリズムを用いてもよい。更にこの多チャネルエコーキャンセル装置を、コンピュータによりプログラムを実行させて機能させることもできる。この場合はこの多チャネルエコーキャンセルプログラムを、CD−ROM、フロッピーディスク、磁気ディスクなどからコンピュータのプログラムメモリにインストールし、あるいは通信回線を通じてダウンロードして、そのプログラムを実行することになる。
【0020】
【発明の効果】
図1に示した実施形態において、マイクロホン16と正面スピーカ12間の距離が約1.5m、スピーカ12と12間の距離が約1.2mの場合、インパルス応答 (k)と (k)のパワー比は壁の反射などにより変わり、1:0.75〜0.85程度となるが、1:0.8とした。この場合の計算機シミュレーションの結果を図3に示す。図3中の実線はこの発明方法による疑似反響路の収束状態を示し、破線は従来法による収束状態を示す。反響路を変化させてから係数誤差(インパルス応答推定値誤差)が−10dBになる時間が、従来法では8.3秒掛ったが、この発明方法によれば5.9秒となり、従来法より約1.5倍の収束速度が得られ、この発明方法が優れていることが確認できた。
【0021】
以上述べたようにこの発明によれば、各チャネルの反響路のインパルス応答に応じてそのチャネルの反響路推定における更新ステップサイズを決定することにより、多チャネルエコーキャンセルの収束速度を向上させることができる。
【図面の簡単な説明】
【図1】この発明の実施形態の機能構成を示す図。
【図2】インパルス応答に応じてステップサイズを決めるための説明図。
【図3】この発明の効果を確認するための、収束状態の計算機シミュレーションを示す図。
【図4】従来の1チャネルエコーキャンセル装置の機能構成を示す図。
【図5】従来の多チャネルエコーキャンセル装置の構成図。
【図6】従来の多チャネルエコーキャンセル装置の具体的機能構成を示す図。
【図7】従来の2チャネルエコーキャンセル装置の機能構成を示す図。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a multi-channel echo canceling method, an apparatus thereof, a program thereof, and a recording medium for canceling a room echo signal that causes a howling and an auditory disturbance in, for example, a communication conference system having a multi-channel playback system. is there.
[0002]
[Prior art]
There is an echo canceling device in order to provide a loudspeaking call system with excellent simultaneous call performance and low echo.
First, an echo canceling method and apparatus configuration of an echo canceling apparatus for one channel will be described with reference to FIG. In a voice call, the received signal obtained by the other party's utterance at the receiving terminal 11 is automatically reproduced from the speaker 12 and automatically sent to the speaker 12 according to the amplitude of the received signal, the power, etc. In some cases, the signal is reproduced from the speaker 12 after some processing is performed on the received signal, such as by adjusting the gain automatically. For this reason, in this specification, the received signal x 1 (k) is not limited to the received signal itself from the other party, and when the received signal is processed, it indicates the processed received signal. Shall. k represents discrete time. The echo canceling device 14 cancels the echo signal y 1 (k) obtained by collecting the received signal x 1 (k) from the speaker 12 through the echo path 15 and collecting into the microphone 16. Here, the reverberation signal y 1 (k) has an impulse response of the reverberation path 15 at time k as h 11 (k, n),
Figure 0003616341
Σ can be modeled as that obtained by a convolution operation such as n = 0 to L-1. L is the number of taps, and is a constant set in advance corresponding to the reverberation time of the echo path 15. First, the received signal accumulation / vector generation unit 17 accumulates the received signal x 1 (k) up to L-1 time past. The accumulated signal is the received signal vector x 1 (k), ie
Figure 0003616341
Is output as However, * T represents transposition of a vector. In the pseudo reverberation signal generation unit 18, the inner product calculation of the received signal vector x 1 (k) of the equation (2) and the pseudo reverberation vector h ^ 11 (k) obtained from the echo path estimation unit 19.
Figure 0003616341
As a result, a pseudo echo signal y ^ 1 (k) is generated. This inner product operation is equivalent to a convolution operation as shown in Equation (1). The echo path estimation unit 19 generates a pseudo echo path vector h 11 (k) used by the pseudo echo signal generation unit 18. The most common algorithm used for echo path estimation is the NLMS algorithm (learning identification method). The NLMS algorithm, the received signal vector x 1 (k) at time k, the residual echo signal e 1 (k), i.e. the estimated echo signal y ^ 1 by the difference circuit 21 from the output y 1 of the microphone 16 (k) (k) Minus the signal
Figure 0003616341
From the above, the pseudo echo path vector h ^ 11 (k + 1) used at time k + 1 is obtained as follows.
Figure 0003616341
However, μ is called a step size parameter and is used for adjustment of the adaptive operation in the range of 0 <μ <2. By repeating the processing as described above, the echo path estimation unit 19 gradually converts the pseudo echo vector h ^ 11 (k) to the time series of the impulse response h 11 (k, n) of the true echo path 15. The echo path vector h 11 (k) as an element, that is,
Figure 0003616341
As a result, the residual echo signal e 1 (k) in the equation (4) can be reduced.
[0003]
In general, in the case of a communication conference system constituted by a reproduction system of N ( > 2) channel and a sound collection system of M ( > 1) channel, the echo cancellation is performed by the configuration shown in FIG. In other words, echo cancellation is performed by connecting N channel echo cancellers 22 1 , 22 2 ,..., 22 M for processing N input 1 output time-series signals between all N channels on the reproduction side and one channel on the sound collection side. This is realized as a system 23. In this case, N × M echo paths 15 nm (1 < n < N, 1 < m < M) exist in the entire system. FIG. 4 shows the N channel echo cancellers 22 1 , 22 2 ,..., 22 M connected between all N channels on the reproduction side and one channel on the sound collection side, which are constituent units of this system. Further, the configuration of the echo canceling device 14 is expanded as shown in FIG. This is the case, for example, of the IEICE Transactions '86 / 10 Vol. J69-A No. 10 “Multi-channel adaptive digital filters”. Here, consider an N-channel echo canceller 22 m whose sound collection side is connected to the m-th sound collection channel (1 < m < M). Since the echo signal collected by the m-th channel microphone 16 m is obtained by adding the reception signals of the respective reproduction channels through the respective echo paths 15 1m to 15 Nm on the sound collection side. It is necessary to devise a method for evaluating the path by uniformly evaluating only the same residual echo signal e m (k) for any reproduction channel. First, with respect to the reception signal of each reproduction channel, the reception signal vector is generated by the reception signal accumulation / vector generation unit (17 1 , 17 2 ,..., 17 N ).
Figure 0003616341
Is generated. Here, L 1 , L 2 ,..., L N are tap numbers, and are constants set in advance corresponding to the reverberation times of the reverberation paths 15 1m , 15 2m ,. These vectors are converted by the vector combiner 24,
Figure 0003616341
Combine with. Also in the echo path estimation unit 19 m , each pseudo echo path vector h ^ 1m (k), h ^ for suspecting N echo paths between each reproduction channel and the m-th sound collection channel. 2m (k), ..., h ^ Nm (k)
Figure 0003616341
Treat as. The update of the pseudo-echo path vector h ^ m (k) is performed using the NLMS algorithm.
Figure 0003616341
It is done like this. In the pseudo echo signal generation unit 18 m , the inner product calculation
Figure 0003616341
Thus, a pseudo echo signal y ^ m (k) for the echo signal y m (k) collected by the mth sound collection channel is generated. Thus, by combining the vectors for each channel and treating them as one vector, the basic processing flow is the same as that of the one-channel echo canceller shown in FIG.
[0004]
[Problems to be solved by the invention]
In the conventional multi-channel echo cancellation method, it is slow that the pseudo echo path almost converges to the true echo path. As a result of research to improve this point, there is a problem in that the update step size μ in the successive estimation of the echo path has the same value in the pseudo echo path for the echo signal for each reproduction signal of the reception signal of each channel. I came to think.
In order to simplify the explanation, as shown in FIG. 7, an example is given in which there are two reception channels and one sound collection channel. Each received signal x 1 (k), x 2 (k) are respectively applied to the channel estimated echo path 18 1, 18 2, channel estimated echo path 18 1, 18 each channel than 2 echo replica y ^ 1 (k ), y ^ 2 (k) are synthesized by the synthesizing circuit 26 is supplied to the overall echo replica y ^ (k) and has been the difference circuit 21, y from the sound collection signal y from the microphone 16 1 (k) ^ (K) is subtracted to obtain a residual echo signal e (k). A residual echo signal e (k), the received signal x 1 (k), x 2 (k) is supplied to the channel echo path estimation unit 19 1, 19 2, channel echo path estimation unit 19 1, 19 2, respectively, The impulse responses h 1 (k) and h 2 (k) of the echo paths 15 1 and 15 2 where the reproduced signals of the received signals x 1 (k) and x 2 (k) reach the microphone 16 1 are sequentially estimated, These estimated impulse responses h ^ 1 (k) and h ^ 2 (k) are set in the channel pseudo echo paths 18 1 and 18 2 , respectively.
[0005]
The estimated channel impulse responses h 1 (k) and h 2 (k) in the channel echo path estimation units 19 1 and 19 2 are as follows from the equation (12). Expression (12) becomes Expression (14) because there are two reception channels.
[0006]
[Expression 1]
Figure 0003616341
[0007]
The echo path 15 1, 15 taps determined by the second reverberation time L 1 and L 2 have the same L.
Since X T (k) X (k) is a scalar in this way, Equation (14) can be written as follows for each row of the matrix.
Figure 0003616341
That is, the channel echo path estimation units 19 1 and 19 2 sequentially estimate the impulse responses h 1 (k) and h 2 (k) of the channel echo paths 15 1 and 15 2 using the equations (15) and (16), respectively. Become. X 1 (k) / (X T (k) X (k)) and X 2 (k) / (X T (k) X (k)) are receptions normalized by the power of reception signals of all channels, respectively. It can be said that it is a signal (vector). X T (k) X (k) is calculated by the power calculation unit 27 and supplied to the estimation units 19 1 and 19 2 . In FIG. 7 , the received signal accumulation vector generation unit in FIG. 6 is necessary but omitted. From equations (15) and (16), the update step size is both equal to μ.
[0008]
However, for example, as shown in FIG. 7, when the speakers 12 1 and 12 2 and the microphones 16 1 and 16 2 are opposed to each other and positioned at each corner of the square, the speaker 12 1 in front of the microphone 16 1 towards echo path 15 1 of the impulse response h 1 (k) is greater energy than the diagonal position of the speaker 12 and second echo path 15 2 from the impulse response h 2 (k). This is not limited to the arrangement of the rectangular corners of the speaker and the microphone, and there is a difference in the energy of the impulse response depending on the echo path in other arrangements, and the same can be said for the multi-channel echo cancellation method of three or more channels. .
[0009]
SUMMARY OF THE INVENTION An object of the present invention is to provide a multi-channel echo canceling method, an apparatus, a program thereof, and a recording medium thereof, which have a faster convergence than conventional methods in consideration of the energy difference of the impulse response.
[0010]
[Means for Solving the Problems]
According to the present invention, the received signal of each channel is applied to the channel pseudo echo path to generate a channel pseudo echo signal, and the channel pseudo echo signal is added to generate a total pseudo echo signal. Is subtracted from the collected sound signal to generate a residual echo signal, and in particular , the input signal magnitude corresponding to each impulse response of the reproduction signal and the collected sound signal of the received signal of each channel is input, and the impulse response of each channel is input. The update step size of each channel is generated so that the relative ratio of the square sum of the coefficients of the series is equal to the relative ratio of the update step size of each channel, the update step size for each channel, each received signal, and the above The impulse response of the echo path of the channel is sequentially estimated from the residual echo signal, and the pseudo echo path of the channel is estimated based on the estimated impulse response. Generated.
[0011]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 shows an embodiment of the present invention. In order to simplify the description, the reception 2 channel and the sound collection 1 channel are taken as an example, and the same reference numerals are given to the portions corresponding to FIG. In this embodiment, signals corresponding to the impulse responses h 1 (k), h 2 (k) of the echo paths 15 1 , 15 2 of each channel are input to the step size generator 31, and the step size for each channel is entered. μ 1 and μ 2 are determined according to the magnitude of the signal corresponding to the impulse response. Signals corresponding to the impulse responses h 1 (k) and h 2 (k) are impulses between the reproduced signals of the received signals x 1 (k) and x 2 (k) and the sound collection signal y (k). It is estimated empirically depending on the response, that is, the impulse responses h 1 (k) and h 2 (k) of the echo paths 15 1 and 15 2 , or the arrangement of the speakers 12 1 and 12 2 and the microphone 16 1. An approximate value or an estimated impulse response set in the channel pseudo echo paths 18 1 and 18 2 is used. When the impulse responses h 1 (k) and h 2 (k) are used, the impulse sound is radiated from only one speaker and the output signal of the microphone 16 is measured in advance.
[0012]
Update step size mu 1, mu 2 is the impulse response h 1 of each channel (k), h 2 (k ) and the corresponding signal (hereinafter h 1 (k), is represented by h 2 (k)) of the size, That is, depending on the amplitude or power, a larger value is set if it is larger. In particular, the update step sizes μ 1 and μ 2 are determined according to the ratio of the impulse responses h 1 (k) and h 2 (k) of each channel to the sum of these powers. In this case, the sum of the update step sizes μ 1 2 and μ 2 2 of these channels is set to 2. For example, it is obtained as follows. Now, assuming that h 1 (k) and h 2 (k) have only one element and these are written on the h plane, as shown in FIG. 2, the update vector h ^ (k) in equation (14) becomes h 1 When the powers of (k) and h 2 (k) are equal, they are on a vector 41 of 45 degrees with respect to the horizontal axis. The lengths of the vectors h 1 (k) and h 2 (k) are expressed as a = √ ( h 1 T (k) h 1 (k)), b = √ ( h 2 T (k) h 2 (k)) Then, the update vector h ^ (k) can be set so as to be on the vector 42 in FIG. 2, so that the pseudo echo path can be converged to a true echo path faster. The step sizes μ 1 and μ 2 are made equal to the ratio of √ (a 2 ) and √ (b 2 ) according to this update vector h ^ (k), that is, μ 1 : μ 2 = √ (a 2 ): √ (b 2 )
And it is sufficient. Therefore, the following simultaneous equations are obtained.
[0013]
μ 1 √ (b 2 ) = μ 2 √ (a 2 )
μ 1 2 + μ 2 2 = 2
Solving these simultaneous equations gives the following.
μ 1 = √ (2a 2 ) / (√ (a 2 + b 2 ))
μ 2 = √ (2b 2 ) / (√ (a 2 + b 2 ))
That is, μ 1 = √ (2 h 1 T (k) h 1 (k)) / (√ (h 1 T (k) h 1 (k)) + √ (h 2 T (k) h 2 (k)) (17)
μ 2 = √ (2 h 2 T (k) h 2 (k)) / (√ (h 1 T (k) h 1 (k)) + √ (h 2 T (k) h 2 (k))) (18)
May be calculated by the step size generation unit 31.
[0014]
Accordingly, the impulse responses h 1 (k) and h 2 (k) are input to the square sum circuits 32 1 and 32 2 , respectively, and the square sum of the elements of each vector.
h 1 T (k) h 1 (k) = Σ i = 0 L−1 h 1 2 (k, i)
h 2 T (k) h 2 (k) = Σ i = 0 L−1 h 2 2 (k, i)
Are calculated, and square sums of these square sums are taken by the square root circuit 33 1 , 33 2 , and a = √ ( h 1 T (k) h 1 (k), b = √ ( h 2 T (k) h 2 (k)) is obtained. these a and b are added together by the addition circuit 34, equation (17), a by the sum of the denominator is obtained. square by dividing circuit 35 1 addition circuit 34 (18) multiplied by 2 in dividing the value computed mu 1 of formula (17) is determined. Similarly multiplication and division circuit 35 2 in by multiplying 2 to a value obtained by dividing the b in the addition value of the adder circuit 34 the formula ( calculation results mu 2 of 18) is obtained.
[0015]
These update step sizes μ 1 and μ 2 are supplied to the corresponding channel echo estimation units 19 1 and 19 2 , respectively. The channel echo path estimation units 19 1 and 19 2 are supplied with received signals x 1 (k) and x 2 (k), respectively, as in the case shown in FIG. The sum of the received signal power of the channels and the residual echo signal e (k) are supplied in common. Channel echo path estimation units 19 1 and 19 2 sequentially perform channel echo path impulse response estimations h 1 (k + 1) and h 2 (k + 1) according to the following equations, respectively.
[0016]
Figure 0003616341
These estimated impulse responses h 1 (k + 1) and h 2 (k + 1) are set in the channel pseudo echo paths 18 1 and 18 2 , respectively. Other processes are the same as those of the apparatus shown in FIG.
[0017]
As described above, the update step sizes μ 1 and μ 2 correspond to the energy ratios of the impulse responses h 1 (k) and h 2 (k) of the echo paths 15 1 and 15 2 , so The step size is increased, and the convergence of the pseudo echo paths 18 1 and 18 2 as a whole is accelerated accordingly.
The step size generator 31 calculates μ 1 (k) and μ 2 (k) at each time k, so that the echo paths 15 1 and 15 2 fluctuate and h 1 (k) and h 2 (k). Even if fluctuates, appropriate μ 1 (k) and μ 2 (k) are set following this. The first step size generating unit 31, the impulse response h 1 (k), also enter the empirical estimate of h 2 (k), then the channel estimated echo path 18 1, 18 2 of the impulse response h By inputting ^ 1 (k), h ^ 2 (k) to the step size generation unit 31, the update step size μ is increased as the channel pseudo echo paths 18 1 and 18 2 converge to the echo paths 15 1 and 15 2. 1 and μ 2 are updated to appropriate values μ 1 (k) and μ 2 (k). Alternatively, the initial values of μ 1 and μ 2 are set to μ 1 = μ 2 and appropriate values used in the prior art are set. Thereafter, h ^ 1 (k) and h ^ 2 (k) are set as step size generation units. 31 may be input to update μ 1 and μ 2 .
[0018]
The present invention can also be applied to a case where there are three or more receiving channels. For example, in the case of N channels, the step size generation unit 31 may perform the following calculation for each channel n (n = 1, 2,..., N).
μ n = √N (h n T (k) h n (k)) / √ (h 1 T (k) h 1 (k) + (h 2 T (k) h 2 (k)) + h N T (k) h N (k))
As understood from this equation, the convergence rate of the pseudo-resonance path of the channel having a large impulse response is determined by the relative size ratio between the channels of the impulse response (amplitude or power) of each channel. What is necessary is just to obtain | require step size so that it may become large. In the above embodiment, the channel echo path estimation unit uses the sum of the powers of received signals of all channels, but this is not always necessary. For example, IEICE Transactions '86 / 10 Vol. J69-A No. As proposed in 10 “Multi-channel adaptive digital filter”, a signal obtained by normalizing the received signal for each channel with its power may be used.
[0019]
The channel echo path estimation algorithm is not limited to the learning identification method, and a projection method or other algorithms may be used. Further, this multi-channel echo canceling apparatus can be made to function by executing a program by a computer. In this case, the multi-channel echo cancellation program is installed in a computer program memory from a CD-ROM, floppy disk, magnetic disk or the like, or downloaded through a communication line, and the program is executed.
[0020]
【The invention's effect】
In the embodiment shown in FIG. 1, the distance between the microphone 16 1 and the front speaker 12 1 of about 1.5 m, when the distance between the speaker 12 1 and 12 2 is about 1.2 m, the impulse response h 1 (k) The power ratio of h 2 (k) varies depending on the reflection of the wall and is about 1: 0.75 to 0.85, but is set to 1: 0.8. The result of the computer simulation in this case is shown in FIG. The solid line in FIG. 3 indicates the convergence state of the pseudo echo path according to the method of the present invention, and the broken line indicates the convergence state according to the conventional method. The time for the coefficient error (impulse response estimated value error) to be −10 dB after changing the echo path took 8.3 seconds with the conventional method, but according to the method of the present invention, it took 5.9 seconds, which is more than the conventional method. A convergence speed of about 1.5 times was obtained, and it was confirmed that the method of the present invention was excellent.
[0021]
As described above, according to the present invention, it is possible to improve the convergence speed of multi-channel echo cancellation by determining the update step size in the echo path estimation of the channel according to the impulse response of the echo path of each channel. it can.
[Brief description of the drawings]
FIG. 1 is a diagram showing a functional configuration of an embodiment of the present invention.
FIG. 2 is an explanatory diagram for determining a step size according to an impulse response.
FIG. 3 is a diagram showing a computer simulation in a converged state for confirming the effect of the present invention.
FIG. 4 is a diagram showing a functional configuration of a conventional one-channel echo cancellation apparatus.
FIG. 5 is a configuration diagram of a conventional multi-channel echo canceling apparatus.
FIG. 6 is a diagram showing a specific functional configuration of a conventional multi-channel echo canceling apparatus.
FIG. 7 is a diagram showing a functional configuration of a conventional two-channel echo cancellation apparatus.

Claims (4)

Nチャネル(Nは2以上の整数)の受話信号をそれぞれ対応するチャネル疑似反響路に印加してチャネル疑似反響信号を生成し、
これらチャネル疑似反響信号を加算して総合疑似反響信号を生成し、
その総合疑似反響信号を、上記各受話信号の再生信号に対する集音信号から差し引いて残留反響信号を生成し、
上記各チャネルの受話信号の再生信号と上記集音信号との間の各インパルス応答を入力して、各チャネルのインパルス応答の級数の係数の二乗和の相対比と各チャネルの更新ステップサイズの相対比が等しくなるように各チャネルの更新ステップサイズを算出し、
上記各チャネルごとに更新ステップサイズと、各チャネルの受話信号と、上記残留反響信号とからそのチャネルの上記再生信号の反響路のインパルス応答を逐次推定し、
その逐次推定したインパルス応答により対応するチャネルの疑似反響路を生成することを特徴とする多チャネルエコーキャンセル方法。
N channel (N is an integer of 2 or more) received signals are respectively applied to corresponding channel pseudo echo paths to generate channel pseudo echo signals,
These channel pseudo echo signals are added to generate an overall pseudo echo signal,
Subtracting the total simulated echo signal from the collected signal for the reproduction signal of each received signal to generate a residual echo signal,
Input each impulse response between the reproduced signal of the received signal of each channel and the collected sound signal, and the relative ratio of the square sum of the coefficient of the series of impulse response of each channel and the update step size of each channel Calculate the update step size for each channel so that the ratios are equal ,
An updating step size for each of the respective channels, and the reception signals of the channels, sequentially estimating the impulse response of the echo path of the reproduction signal of the channel from the said residual echo signal,
A multi-channel echo cancellation method, wherein a pseudo echo path of a corresponding channel is generated based on the sequentially estimated impulse response.
Nチャネル(Nは2以上の整数)の受話信号がそれぞれ印加されて、チャネル疑似反響信号を生成するN個のチャネル疑似反響路と、
上記N個のチャネル疑似反響信号を加算して総合疑似信号を生成する合成回路と、
上記N個の受話信号の再生信号に対する集音信号から上記総合疑似信号を差し引いて残留反響信号を生成する減算回路と、
上記N個の受話信号の再生信号と上記集音信号との間の各インパルス応答を入力して、各チャネルのインパルス応答の級数の係数の二乗和の相対比と各チャネルの更新ステップサイズの相対比が等しくなるようにN個の更新ステップサイズを求めるステップサイズ生成部と、
上記各チャネルごとに上記更新ステップサイズと、各チャネルの受話信号と、上記残留反響信号とが入力され、そのチャネルの上記再生信号の反響路のインパルス応答を逐次推定し、その推定したインパルス応答を対応する上記チャネル疑似反響路に設定するN個の反響路推定部と、
を具備する多チャネルエコーキャンセル装置。
N channels (N is an integer of 2 or more) of received signals are applied to each of the N channel pseudo echo paths to generate a channel pseudo echo signal;
A synthesis circuit for generating a total pseudo signal by adding the N channel pseudo echo signals;
A subtracting circuit that generates a residual echo signal by subtracting the synthetic pseudo signal from the collected sound signal for the reproduced signals of the N received signals;
Each impulse response between the reproduced signal of the N received signals and the collected sound signal is input, and the relative ratio of the square sum of the coefficients of the series of impulse responses of each channel and the relative update step size of each channel A step size generator for obtaining N update step sizes so that the ratios are equal ;
The update step size, the received signal of each channel, and the residual echo signal are input for each channel, and the impulse response of the echo path of the reproduced signal of that channel is sequentially estimated, and the estimated impulse response is N echo path estimators set in the corresponding channel pseudo echo path;
A multi-channel echo canceling apparatus.
請求項1記載の多チャネルエコーキャンセル方法をコンピュータに実行させるプログラム。Program for executing a multi-channel echo canceling method of claim 1 Symbol placement on the computer. 請求項記載のプログラムを記録したコンピュータ読み出し可能な記録媒体。A computer-readable recording medium on which the program according to claim 3 is recorded.
JP2001051387A 2001-02-27 2001-02-27 Multi-channel echo cancellation method, apparatus thereof, program thereof, and recording medium Expired - Fee Related JP3616341B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2001051387A JP3616341B2 (en) 2001-02-27 2001-02-27 Multi-channel echo cancellation method, apparatus thereof, program thereof, and recording medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2001051387A JP3616341B2 (en) 2001-02-27 2001-02-27 Multi-channel echo cancellation method, apparatus thereof, program thereof, and recording medium

Publications (2)

Publication Number Publication Date
JP2002261659A JP2002261659A (en) 2002-09-13
JP3616341B2 true JP3616341B2 (en) 2005-02-02

Family

ID=18912179

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2001051387A Expired - Fee Related JP3616341B2 (en) 2001-02-27 2001-02-27 Multi-channel echo cancellation method, apparatus thereof, program thereof, and recording medium

Country Status (1)

Country Link
JP (1) JP3616341B2 (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4504891B2 (en) * 2005-08-31 2010-07-14 日本電信電話株式会社 Echo canceling method, echo canceling apparatus, program, recording medium
JP4504892B2 (en) * 2005-08-31 2010-07-14 日本電信電話株式会社 Echo canceling method, echo canceling apparatus, program, recording medium
JP4852068B2 (en) * 2008-05-20 2012-01-11 日本電信電話株式会社 Stereo acoustic echo cancellation method, stereo acoustic echo cancellation apparatus, stereo acoustic echo cancellation program, and recording medium thereof
US8135140B2 (en) * 2008-11-20 2012-03-13 Harman International Industries, Incorporated System for active noise control with audio signal compensation
JP5264687B2 (en) * 2009-11-27 2013-08-14 日本電信電話株式会社 Echo canceling method, echo canceling device, echo canceling program

Also Published As

Publication number Publication date
JP2002261659A (en) 2002-09-13

Similar Documents

Publication Publication Date Title
JP3654470B2 (en) Echo canceling method for subband multi-channel audio communication conference
JP3397269B2 (en) Multi-channel echo cancellation method
US5774562A (en) Method and apparatus for dereverberation
US7693291B2 (en) Multi-channel frequency-domain adaptive filter method and apparatus
JP2004349806A (en) Multichannel acoustic echo canceling method, apparatus thereof, program thereof, and recording medium thereof
US6694020B1 (en) Frequency domain stereophonic acoustic echo canceller utilizing non-linear transformations
JP3616341B2 (en) Multi-channel echo cancellation method, apparatus thereof, program thereof, and recording medium
JP3756828B2 (en) Reverberation elimination method, apparatus for implementing this method, program, and recording medium therefor
JP3864914B2 (en) Echo suppression device
JP3673727B2 (en) Reverberation elimination method, apparatus thereof, program thereof, and recording medium thereof
JP4041770B2 (en) Acoustic echo cancellation method, apparatus, program, and recording medium
JP3628267B2 (en) Multi-channel echo cancellation method, apparatus thereof, program thereof and recording medium thereof
JP3402427B2 (en) Multi-channel echo cancellation method and apparatus
JP7373947B2 (en) Acoustic echo cancellation device, acoustic echo cancellation method and acoustic echo cancellation program
JPH09261135A (en) Acoustic echo erasion device
CN112863532A (en) Echo suppressing device, echo suppressing method, and storage medium
JP4852068B2 (en) Stereo acoustic echo cancellation method, stereo acoustic echo cancellation apparatus, stereo acoustic echo cancellation program, and recording medium thereof
JP6075783B2 (en) Echo canceling apparatus, echo canceling method and program
EP4016977A1 (en) Apparatus and method for filtered-reference acoustic echo cancellation
JP3339612B2 (en) Echo cancellation method for multi-channel audio communication conference
JP3700838B2 (en) Multi-channel echo cancellation method, multi-channel echo cancellation device, multi-channel echo cancellation program
JP6356087B2 (en) Echo canceling apparatus, method and program
EP1351479A1 (en) Generating an undisturbed signal out of an audio signal including a disturbing signal
Tahir et al. On implementation of prediction error method-based Adaptive Feedback Cancellation (AFC) in digital hearing aids
Nakagawa Control of feedback for assistive listening devices

Legal Events

Date Code Title Description
A977 Report on retrieval

Free format text: JAPANESE INTERMEDIATE CODE: A971007

Effective date: 20040601

A131 Notification of reasons for refusal

Free format text: JAPANESE INTERMEDIATE CODE: A131

Effective date: 20040713

A521 Written amendment

Free format text: JAPANESE INTERMEDIATE CODE: A523

Effective date: 20040910

TRDD Decision of grant or rejection written
A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

Effective date: 20041012

A61 First payment of annual fees (during grant procedure)

Free format text: JAPANESE INTERMEDIATE CODE: A61

Effective date: 20041104

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20071112

Year of fee payment: 3

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20081112

Year of fee payment: 4

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20091112

Year of fee payment: 5

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20101112

Year of fee payment: 6

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20101112

Year of fee payment: 6

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20111112

Year of fee payment: 7

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20111112

Year of fee payment: 7

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20121112

Year of fee payment: 8

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20121112

Year of fee payment: 8

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20131112

Year of fee payment: 9

LAPS Cancellation because of no payment of annual fees