JP2006148540A - Device and method for calculating call state value - Google Patents

Device and method for calculating call state value Download PDF

Info

Publication number
JP2006148540A
JP2006148540A JP2004335900A JP2004335900A JP2006148540A JP 2006148540 A JP2006148540 A JP 2006148540A JP 2004335900 A JP2004335900 A JP 2004335900A JP 2004335900 A JP2004335900 A JP 2004335900A JP 2006148540 A JP2006148540 A JP 2006148540A
Authority
JP
Japan
Prior art keywords
echo cancellation
signal
value
call state
state value
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.)
Granted
Application number
JP2004335900A
Other languages
Japanese (ja)
Other versions
JP4452162B2 (en
Inventor
Katsuhiro Fukui
勝宏 福井
Suehiro Shimauchi
末廣 島内
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 JP2004335900A priority Critical patent/JP4452162B2/en
Publication of JP2006148540A publication Critical patent/JP2006148540A/en
Application granted granted Critical
Publication of JP4452162B2 publication Critical patent/JP4452162B2/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

<P>PROBLEM TO BE SOLVED: To highly accurately and stably detect a call state by utilizing a device for calculating a call state value, which is capable of suppressing deterioration of performance of an echo canceler due to conventional problems, high in detection accuracy of double-talk, and less dependent on an input signal, and also, utilizing a method for calculating the call state value. <P>SOLUTION: The device for calculating a speech state value is composed of an echo cancellation amount estimator (20), which integrates the ratio of power peak hold values of each of an echo canceling signal e(k) and a sound pickup signal y(k) at time point k extending over to a fixed number L of samples, and calculates an echo cancellation amount estimate value b(k), on the basis of the integrated value and a speech state value c(k); and a call-state calculator (30) which integrates the ratio of power peak hold values of each of an echo canceling signal and a sound pickup signal at the time point (k+1) over the fixed number L of the samples, and calculates the call state value c(k), on the basis of the integrated value and the echo cancellation amount estimate value b(k). <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

本発明は、ハンズフリー通話、遠隔会議、などにおけるエコーキャンセラを使用した系における通話状態を数値で表した通話状態値を算出する装置およびその方法に関するものである。   The present invention relates to an apparatus and a method for calculating a call state value in which a call state in a system using an echo canceller in a hands-free call, a remote conference, and the like is expressed numerically.

図1に示すように、端子9に入力した受話信号x(k)がスピーカ11からマイクロホン12ヘ回り込むエコーを消去する従来のエコーキャンセラ10は、スピーカ11とマイクロホン12間のエコー経路のインパルス応答を要素として持つ長さ(即ちタップ数)Lのベクトルhの擬似特性h'(k)を保持する擬似エコー経路を適応フィルタ13で実現している。ここで、kは、所定間隔の離散時間を指すステップ数である。なお、スピーカ11に与える信号、マイクロホン12で収音された信号はアナログ信号であり、以下の説明では、ディジタル信号を扱うので、それぞれDA変換器、AD変換器によって変換を行う必要があるが、それらは当然のことであり、図示してない。また、エコーキャンセラ10の信号処理技術自体は周知であり、各種信号を表す記号がベクトルであるか否かを区別して表してなくてもこの技術分野の技術者であれば容易に理解できるので、本明細書の文章中においては信号の記号がベクトルであるか否かの区別をしない。ただし、数式イメージ中および図面中にはベクトルを太字で表している。   As shown in FIG. 1, a conventional echo canceller 10 that eliminates an echo that a reception signal x (k) input to a terminal 9 wraps around from a speaker 11 to a microphone 12 has an impulse response of an echo path between the speaker 11 and the microphone 12. A pseudo echo path that holds the pseudo characteristic h ′ (k) of the vector h of length (ie, the number of taps) L as an element is realized by the adaptive filter 13. Here, k is the number of steps indicating discrete time at a predetermined interval. Note that the signal given to the speaker 11 and the signal collected by the microphone 12 are analog signals, and in the following description, since digital signals are handled, it is necessary to perform conversion by a DA converter and an AD converter, respectively. They are of course not shown. In addition, the signal processing technique itself of the echo canceller 10 is well known, and even a technician in this technical field can easily understand even if the symbols representing various signals are not distinguished from each other. In the text of this specification, it is not distinguished whether the symbol of the signal is a vector. However, the vector is shown in bold in the mathematical image and in the drawings.

適応フィルタ13は受話信号x(k)と擬似特性h'(k)との畳み込み演算により擬似エコー信号d'(k)を生成し、実際のエコー信号を含むマイクロホン12の収音信号(エコー消去前信号とも呼ぶ)y(k)から減算器15により減算することで、エコーキャンセラの出力信号でもあるエコー消去信号(送話信号とも呼ぶ)e(k)を出力する。また、係数更新部14は受話信号x(k)と誤差信号e(k)とを用いて、擬似特性h'(k)の特性を随時推定し、適応フィルタ13に設定する。
係数更新部14による、例えば学習同定アルゴリズムを用いた場合の擬似特性h'(k)の推定は、

Figure 2006148540
The adaptive filter 13 generates a pseudo echo signal d ′ (k) by a convolution operation between the received signal x (k) and the pseudo characteristic h ′ (k), and collects a sound signal (echo cancellation) of the microphone 12 including the actual echo signal. By subtracting from the subtractor 15 from y (k) (also referred to as the previous signal), an echo cancellation signal (also referred to as a transmission signal) e (k) that is also an output signal of the echo canceller is output. Further, the coefficient updating unit 14 estimates the characteristic of the pseudo characteristic h ′ (k) as needed using the received signal x (k) and the error signal e (k), and sets the characteristic in the adaptive filter 13.
The estimation of the pseudo characteristic h ′ (k) when the coefficient updating unit 14 uses a learning identification algorithm, for example,
Figure 2006148540

で表される。ここで、x(k)=[x(k),x(k-1),…,x(k-L+1)]T、αは係数の更新幅を与えるステップサイズであり、0から2の間の値をとる実数である。δは分母が0になることを防止するための微小な定数である。式(1) が示すように、前回の擬似特性h’(k-1)に対し更新量を加えて今回の擬似特性h’(k)を得る。
エコーキャンセラは、マイクロホン12で収音した収音信号y(k)に、スピーカ11から放射された受話信号x(k)のみが含まれる状態(シングルトーク)のときに正常に動作する。実用上の問題として、エコー信号に送話者の信号が混入する同時通話、いわゆるダブルトークが頻繁に起こり、係数更新部14において適応フィルタ13の学習適応制御が誤動作し、エコーの消し残りが目立つようになる。このため、ダブルトーク検出部16はエコー消去信号e(k)と収音信号y(k)を使ってダブルトークを検出すると、係数更新部14での適応フィルタ13のタップ係数の更新を停止させ、エコー消去量が劣化しないようにしている。
It is represented by Here, x (k) = [x (k), x (k-1),..., X (k-L + 1)] T , α is a step size that gives a coefficient update width, and is 0 to 2 A real number that takes a value between. δ is a minute constant for preventing the denominator from becoming zero. As shown in Expression (1), the update amount is added to the previous pseudo characteristic h ′ (k−1) to obtain the current pseudo characteristic h ′ (k).
The echo canceller operates normally when the collected signal y (k) collected by the microphone 12 includes only the received signal x (k) radiated from the speaker 11 (single talk). As a practical problem, simultaneous calls in which the sender's signal is mixed into the echo signal, so-called double talk, frequently occur, the learning adaptive control of the adaptive filter 13 malfunctions in the coefficient update unit 14, and the remaining echo is not noticeable. It becomes like this. For this reason, when the double talk detecting unit 16 detects double talk using the echo cancellation signal e (k) and the collected sound signal y (k), the coefficient updating unit 14 stops updating the tap coefficient of the adaptive filter 13. The echo cancellation amount is prevented from deteriorating.

ダブルトーク検出部16における、ダブルトークの検出は、特許文献1に記載の手法を用いると、次式

Figure 2006148540
The detection of double talk in the double talk detector 16 uses the method described in Patent Document 1 and
Figure 2006148540

によって行う。ここでKは、エコー消去信号e(k)と収音信号y(k)の対数振幅比に基づいてダブルトークを判定するための予め決めた閾値、qは整数係数であり、いずれも適当な値に設定できる。式(2)の対数振幅比はエコー消去量に対応(正確にはエコー消去量の逆数に対応)している。また、式(3)、式(4)のパワー計算は、エコー除去前のパワーY(k-1), E(k-1)と、エコー除去後の信号y(k), e(k)のパワーのそれぞれ一つ前の値を使用して、巡回型フィルタ演算を行うものである。
特公平7-40680.
To do. Here, K is a predetermined threshold value for determining double talk based on the logarithmic amplitude ratio of the echo cancellation signal e (k) and the collected sound signal y (k), and q is an integer coefficient, both of which are appropriate. Can be set to a value. The logarithmic amplitude ratio in Equation (2) corresponds to the echo cancellation amount (more precisely, the inverse of the echo cancellation amount). In addition, the power calculations of Equation (3) and Equation (4) are based on the power Y (k-1), E (k-1) before echo removal and the signals y (k), e (k) after echo removal. A cyclic filter operation is performed using the previous value of each of the powers.
7-40680.

しかしながら、上記の式(2)において、信号波形の各離散時間の振幅値による振幅比の誤差が大きいため、ダブルトーク検出精度が劣化するという問題点があった。また、振幅比を予め決めた閾値Kと比較して決定するダブルトークの検出精度は、入力信号によってエコー消去量が異なるため、入力信号に大きく依存するという問題点もあった。
本発明の課題は、このような問題に起因するエコーキャンセラの性能劣化を抑圧し、ダブルトーク検出精度が高く、入力信号への依存が小さい通話状態値算出装置および算出方法を提供することである。
However, in the above equation (2), there is a problem that the accuracy of double talk detection deteriorates because the error of the amplitude ratio due to the amplitude value of each discrete time of the signal waveform is large. In addition, the detection accuracy of double talk, which is determined by comparing the amplitude ratio with a predetermined threshold value K, has a problem that it greatly depends on the input signal because the echo cancellation amount differs depending on the input signal.
An object of the present invention is to provide a call state value calculation apparatus and calculation method that suppresses performance degradation of an echo canceller due to such a problem, has high doubletalk detection accuracy, and is less dependent on an input signal. .

この発明によれば、一連の離散的時点の時点kにおけるエコー消去信号とエコー消去前信号のそれぞれ一定サンプル数Lに渡るピークホールドの比の積算値と、与えられた通話状態値とからエコー消去量推定値を算出し、時点(k+1)におけるエコー消去信号とエコー消去前信号のそれぞれ一定サンプル数Lに渡るピークホールドの比の積算値と、上記エコー消去量推定値とから上記通話状態値を算出する。   According to the present invention, echo cancellation is performed from the integrated value of the ratio of peak hold over a certain number of samples L of the echo cancellation signal and the signal before echo cancellation at a time point k in a series of discrete time points, and a given call state value. The amount of estimated value is calculated, and the call state is calculated from the integrated value of the ratio of peak hold over a fixed number of samples L of the echo cancellation signal and the signal before echo cancellation at time (k + 1) and the echo cancellation amount estimation value. Calculate the value.

この発明によれば、エコー消去前信号とエコー消去信号のそれぞれ一定サンプル数に渡るピークホールドの比の積算値とエコー消去量推定値に基づいて通話状態値を算出しているので、従来のような各離散時間の振幅比の誤差を避けることができる。また、通話状態値を算出するためのエコー消去量推定値はエコー消去信号e(k)とエコー消去前信号y(k)から求めることができるため、ダブルトーク検出精度が受話信号に大きく依存するという従来の問題を解決できる。   According to the present invention, since the call state value is calculated based on the integrated value of the ratio of peak hold over the predetermined number of samples of the pre-echo cancellation signal and the echo cancellation signal and the estimated echo cancellation amount, Such an error in the amplitude ratio of each discrete time can be avoided. In addition, since the echo cancellation amount estimated value for calculating the call state value can be obtained from the echo cancellation signal e (k) and the signal y (k) before echo cancellation, the double talk detection accuracy greatly depends on the received signal. Can solve the conventional problem.

実施例1
図2を参照して本発明による通話状態値算出装置をエコーキャンセラ10に適用した実施例を以下に説明する。即ち、入力受話信号x(k)を適応等化フィルタ13により処理して擬似エコー信号d’(k)を生成し、マイクロホン12で収音された収音信号(エコー消去前信号)y(k)から擬似エコー信号d’(k)を減算器15で減算してエコー消去信号(送話信号)e(k)が生成される。係数更新部14はエコー消去信号e(k)と受話信号x(k)とに基づいて式(1) と同様のタップ係数更新を行うが、この発明による通話状態値算出装置100によって算出された通話状態値c(k)により係数更新14における係数更新のステップ係数αを制御する。
Example 1
An embodiment in which the call state value calculation apparatus according to the present invention is applied to the echo canceller 10 will be described below with reference to FIG. That is, the input received signal x (k) is processed by the adaptive equalization filter 13 to generate a pseudo echo signal d ′ (k), and the collected sound signal (the signal before echo cancellation) y (k) collected by the microphone 12. ) To subtract the pseudo echo signal d ′ (k) by the subtracter 15 to generate an echo cancellation signal (transmission signal) e (k). The coefficient updating unit 14 performs tap coefficient updating similar to the equation (1) based on the echo cancellation signal e (k) and the received signal x (k), but is calculated by the call state value calculating apparatus 100 according to the present invention. The coefficient update step coefficient α in the coefficient update 14 is controlled by the call state value c (k).

図2の構成では、通話状態値算出装置100はエコー消去量推定部20と、通話状態値算出部30により構成されている。通話状態値算出部30は収音信号y(k)とエコー消去信号e(k)とエコー消去量推定値b(k)が入力され、通話状態値c(k)を出力する。エコー消去量推定部20は収音信号y(k)とエコー消去信号e(k)と通話状態値c(k)とが入力され、エコー消去量推定値b(k)を出力する。通話状態値c(k)は、マイクロホン12によりエコーのみ収音されるシングルトーク状態であると推定された信頼度を示す数値で、最大値εに近いほど信頼度が高くシングルトーク状態であることを示し、ゼロに近いほど信頼度が低くシングルトークではない状態であることを示す。   In the configuration of FIG. 2, the call state value calculation device 100 includes an echo cancellation amount estimation unit 20 and a call state value calculation unit 30. The call state value calculation unit 30 receives the sound pickup signal y (k), the echo cancellation signal e (k), and the echo cancellation amount estimated value b (k), and outputs the call state value c (k). The echo cancellation amount estimation unit 20 receives the collected sound signal y (k), the echo cancellation signal e (k), and the call state value c (k), and outputs an echo cancellation amount estimation value b (k). The call state value c (k) is a numerical value indicating the reliability estimated to be a single talk state in which only echoes are picked up by the microphone 12. The closer the maximum value ε, the higher the reliability and the single talk state. The closer to zero, the lower the reliability and the non-single talk state.

通話状態値算出部30は通話状態値c(k)を次式により算出する。

Figure 2006148540
min{a,b}はa, bのうち値の小さいほうを選択することを表し、選択した値を通話状態値c(k)とする。εは通話状態値c(k)の上限を決定する係数であり、適当な正の値に設定できる。PH{}はピークホールドを表す。
エコー消去量推定部20は、式(5)に用いられるエコー消去量推定値b(k)を以下のように算出する。
Figure 2006148540
The call state value calculation unit 30 calculates the call state value c (k) by the following equation.
Figure 2006148540
min {a, b} indicates that the smaller one of a and b is selected, and the selected value is set as a call state value c (k). ε is a coefficient that determines the upper limit of the call state value c (k), and can be set to an appropriate positive value. PH {} represents peak hold.
The echo cancellation amount estimation unit 20 calculates the echo cancellation amount estimation value b (k) used in Equation (5) as follows.
Figure 2006148540

βはエコー消去量推定値を徐々に減少させる係数であり、γはεより小さい予め決めた正の値であり、エコー消去量推定値b(k)を更新させるか否かを判定する閾値である。c(k)がγ以下の場合はシングルトークでない可能性が高いので、エコー消去信号e(k)と収音信号y(k)を使ってのエコー消去量推定値の計算を行わず、前回の時点(k-1)でのエコー消去量推定値b(k-1)に係数βを乗算した値を今回のエコー消去量推定値b(k)として使用する。c(k)がγより大であればシングルトークの可能性が高いので、エコー消去信号e(k)と収音信号y(k)から式(6)のようにエコー消去量推定値b(k)を算出する。 β is a coefficient for gradually decreasing the echo cancellation amount estimated value, γ is a predetermined positive value smaller than ε, and is a threshold value for determining whether or not to update the echo cancellation amount estimated value b (k). is there. If c (k) is less than γ, there is a high possibility that it is not a single talk.Therefore, the echo cancellation amount estimated value is not calculated using the echo cancellation signal e (k) and the collected sound signal y (k). A value obtained by multiplying the estimated echo cancellation amount b (k−1) at the time (k−1) by the coefficient β is used as the current echo cancellation amount estimation value b (k). If c (k) is larger than γ, there is a high possibility of single talk, so the echo cancellation amount estimate b () is calculated from the echo cancellation signal e (k) and the collected sound signal y (k) as shown in Equation (6). k) is calculated.

通話状態値c(k)は、エコー消去量推定値b(k)と、収音信号y(k)とエコー消去信号e(k)のピークホールド値の比との関係により、通話状態として、マイクロホンにエコーのみ収音されるシングルトーク状態とシングルトークではない状態の2つの状態の連続値を返す。エコーキャンセラが十分に収束していると仮定すれば、通話状態値c(k)は常に正の値となり、シングルトーク状態である信頼度が高いほど係数εに近い値を、シングルトークでない状態である信頼度が高いほど0に近い値を算出し、適応フィルタヘのフィードバックだけでなく、制御情報として様々なエコーキャンセラのパラメータに適用される。   The call state value c (k) is a call state based on the relationship between the echo cancellation amount estimated value b (k) and the ratio of the peak hold value of the collected sound signal y (k) and the echo cancellation signal e (k). Returns a continuous value of two states, a single talk state where only echo is picked up by the microphone and a non-single talk state. Assuming that the echo canceller has sufficiently converged, the call state value c (k) is always a positive value, and the higher the reliability in the single talk state, the closer to the coefficient ε, in the non-single talk state. As the certain reliability increases, a value closer to 0 is calculated and applied not only to feedback to the adaptive filter but also to various echo canceller parameters as control information.

図3は式(5), (6)を計算する通話状態値算出部30とエコー消去量推定部20の構成例を示し、図4はその計算を行う処理手順を示す。式(5)と(6)を比較すると、式(5)中のΣによる総和を求める式は式(6)中のΣによる総和を求める式と同じである。ただし、前者が時点kにおける最新のL個のサンプルについての総和であるのに対し、後者は時点(k-1)における最新のL個のサンプルについての総和である。したがって、ここでは総和Σの計算において常に1時点前の総和Σを保持しておくことにより、重複演算を避ける構成を示す。また、総和中のピークホールド値PH{(10log10({e(k-i)}2+δ)}の具体的な求め方は、式の表現を簡略化するため

Figure 2006148540
とおくと、次式
Figure 2006148540
FIG. 3 shows a configuration example of the call state value calculation unit 30 and the echo cancellation amount estimation unit 20 for calculating the equations (5) and (6), and FIG. 4 shows a processing procedure for performing the calculation. Comparing equations (5) and (6), the equation for obtaining the sum by Σ in equation (5) is the same as the equation for obtaining the sum by Σ in equation (6). However, the former is the sum for the latest L samples at time point k, while the latter is the sum for the latest L samples at time point (k−1). Therefore, here, a configuration is shown in which, in the calculation of the total sum Σ, the sum Σ one point before is always held, thereby avoiding duplicate calculation. In addition, a specific method for obtaining the peak hold value PH {(10log 10 ({e (ki)} 2 + δ)} in the sum total is to simplify the expression.
Figure 2006148540
The following formula
Figure 2006148540

により求める。max{a, b}はa,bのうち、値の大きい方を選択することを表す。τは0より大で1より小の予め決めた値である。
図3に示すように、エコー消去量推定部20は、エコー消去信号e(k)および収音信号y(k)についてのパワーピークホールド値PH{E(k-i)}、PH{Y(k-i)}を算出するピークホールド算出部21E,21Yと、それらのピークホールド値の差分(対数パワー比に対応)を求める減算器22Aと、減算結果、即ちパワーピーク比をi=0,…,L-1について積算する総和部22Bと、比の積算値に1/Lを乗算する(即ちLで割り算する)ことにより平均化し、暫定エコー消去量推定値b’(k)とする乗算器23と、暫定エコー消去量推定値b’(k)をホールドする暫定推定値ホールド部24と、暫定推定値ホールド部24に保持されている前回の暫定エコー消去量推定値b’(k-1)に係数βを乗算する乗算器25と、その乗算結果βb’(k-1)と暫定エコー消去量推定値b’(k)が与えられ、c(k)γであればβb’(k-1)を、そうでなければb’(k)をエコー消去量推定値b(k)として出力する推定値選択部26と、推定値選択部26が出力したエコー消去量推定値b(k)ホールドする推定値ホールド部27とから構成されている。減算器22Aと総和部22Bはピーク比積算部22を構成している。
Ask for. max {a, b} represents selecting the larger one of a and b. τ is a predetermined value larger than 0 and smaller than 1.
As shown in FIG. 3, the echo cancellation amount estimation unit 20 includes power peak hold values PH {E (ki)} and PH {Y (ki) for the echo cancellation signal e (k) and the collected sound signal y (k). }, And a subtractor 22A for obtaining a difference (corresponding to a logarithmic power ratio) between the peak hold values, and a subtraction result, that is, a power peak ratio is set to i = 0,. A summation unit 22B that accumulates 1 and a multiplier 23 that averages by multiplying the integrated value of the ratio by 1 / L (that is, divides by L) to obtain a provisional echo cancellation amount estimated value b ′ (k); A provisional estimated value hold unit 24 that holds the provisional echo cancellation amount estimated value b ′ (k), and a coefficient for the previous provisional echo cancellation amount estimated value b ′ (k−1) held in the temporary estimated value hold unit 24 a multiplier 25 for multiplying the beta, the multiplication result βb '(k-1) and the tentative echo cancellation amount estimated value b' (k) is given, c (k) 'The (k-1), otherwise b' if gamma .beta.b estimated value selection section 26 for outputting a (k) of the echo cancellation amount estimate b (k), the estimated value selection unit 26 is output An estimated value hold unit 27 for holding an echo cancellation amount estimated value b (k). The subtractor 22 </ b> A and the summation unit 22 </ b> B constitute a peak ratio integration unit 22.

ピークホールド算出部21Eは最新のL個のエコー消去信号サンプルe(k), e(k-1), ..., e(k-L+1)を保持する記憶部21E1と、式(7)による時点(k-i)の対数パワーE(k-1)を計算するサンプルパワー算出部21E2と、対数パワーE(k-1)に係数τを乗算する乗算器21E3と、時点 (k-i-1)のピークホールド値PH{E(k-i-1)}とτE(k-i)のうち、値の大きい方をピークホールド値PH{E(k-i)}として選択出力するピーク値選択部21E4と、選択したピークホールド値PH{E(k-i)}を保持するピークホールド部21E5とから構成されている。
収音信号y(k)に対するピークホールド算出部21Yもピークホールド算出部21Eと同様に構成されており、説明を省略する。
The peak hold calculation unit 21E includes a storage unit 21E1 that holds the latest L echo cancellation signal samples e (k), e (k−1),..., E (k−L + 1), and an expression (7 ), The sample power calculation unit 21E2 that calculates the logarithmic power E (k-1) at the time point (ki), the multiplier 21E3 that multiplies the logarithmic power E (k-1) by the coefficient τ, and the time point (ki-1). Of the peak hold value PH {E (ki-1)} and τE (ki), the peak value selection unit 21E4 that selectively outputs the larger value as the peak hold value PH {E (ki)}, and the selected peak And a peak hold unit 21E5 that holds a hold value PH {E (ki)}.
The peak hold calculation unit 21Y for the collected sound signal y (k) is also configured in the same manner as the peak hold calculation unit 21E, and a description thereof will be omitted.

一方、通話状態値算出部30は、エコー消去量推定部20からの暫定エコー消去量推定値b’(k)と、推定値ホールド部27からの前時点(k-1)でのエコー消去量推定値b(k-1)との比rB=b’(k)/{b(k-1)+τ}を求める比算出部31と、算出された比rBと所定値εのうち、値の小さい方を通話状態値c(k)として出力すると共に、エコー消去量推定部20の推定値選択部26に与える通話状態値選択部32とから構成されている。
図4の、図3で説明したエコー消去量推定部20と通話状態値算出部30から構成されたこの発明の通話状態値算出装置100による通話状態値c(k)生成の処理手順を示す。
On the other hand, the call state value calculation unit 30 calculates the provisional echo cancellation amount b ′ (k) from the echo cancellation amount estimation unit 20 and the echo cancellation amount at the previous time point (k−1) from the estimation value hold unit 27. A ratio calculation unit 31 for obtaining a ratio r B = b ′ (k) / {b (k−1) + τ} with the estimated value b (k−1), and among the calculated ratio r B and the predetermined value ε The lower one is output as the call state value c (k), and the call state value selection unit 32 is provided to the estimated value selection unit 26 of the echo cancellation amount estimation unit 20.
FIG. 4 shows a processing procedure for generating a call state value c (k) by the call state value calculating apparatus 100 according to the present invention, which includes the echo cancellation amount estimating unit 20 and the call state value calculating unit 30 described in FIG.

ステップS1:通話状態値c(k)の初期値をεに設定する。これは、初期状態はシングルトークであることを仮定している。この初期値の設定は通話状態値算出部30の通話状態値選択部32において、c(k)として無条件にεを選択出力することである。
ステップS2:エコー消去信号の最近のL個のサンプルe(k), e(k-1), ..., e(k-L+1)と収音信号の最近のL個のサンプルy(k), y(k-1), ..., y(k-L+1)をそれぞれ記憶部21E1, 21Y1に保持する。
ステップS3:エコー消去信号のサンプルe(k-i)のパワーE(k-i)と、収音信号のサンプルy(k-i)のパワーY(k-i)をサンプルパワー算出部21E2,21Y2により算出する。
Step S1: The initial value of the call state value c (k) is set to ε. This assumes that the initial state is single talk. The setting of the initial value is that the call state value selection unit 32 of the call state value calculation unit 30 selects and outputs ε unconditionally as c (k).
Step S2: The latest L samples e (k), e (k-1), ..., e (k-L + 1) of the echo cancellation signal and the latest L samples y ( k), y (k−1),..., y (k−L + 1) are stored in the storage units 21E1 and 21Y1, respectively.
Step S3: The sample power calculators 21E2 and 21Y2 calculate the power E (ki) of the sample e (ki) of the echo cancellation signal and the power Y (ki) of the sample y (ki) of the collected sound signal.

ステップS4:エコー消去信号のサンプルパワーのピークホールド値PH{E(k-i)}と、収音信号のサンプルパワーのピークホールド値PH{Y(k-i)}を求める。
ステップS5:エコー消去信号および収音信号のピークホールド値の比を求め、そのピークホールド値の比をi=0,…,L-1について積算し、積算結果をLで除算して暫定エコー消去量推定値b’(k)として求める。
ステップS6:現時点kの暫定エコー消去量推定値b’(k)と直前の時点(k-1)のエコー消去量推定値b(k-1)とから比rB=b’(k)/{b(k-1)+δ}を求める。
Step S4: Obtain the peak hold value PH {E (ki)} of the sample power of the echo cancellation signal and the peak hold value PH {Y (ki)} of the sample power of the collected sound signal.
Step S5: The ratio of the peak hold value of the echo cancellation signal and the collected sound signal is obtained, the ratio of the peak hold value is integrated for i = 0,..., L-1, and the provisional echo cancellation is performed by dividing the integration result by L. It is obtained as a quantity estimate b ′ (k).
Step S6: The ratio r B = b ′ (k) / from the provisional echo cancellation amount estimated value b ′ (k) at the current time k and the echo cancellation amount estimated value b (k−1) at the immediately previous time point (k−1). {B (k-1) + δ} is obtained.

ステップS7:所定値εと比rBのうち、値の小さい方を通話状態値c(k)として出力するとともに、エコー消去量推定部20の推定値選択部26に与える。前述のように、初回では無条件でc(k)=εとされる。
ステップS8:通話状態値c(k)がγ以下か判定する。
ステップS9:c(k)がγ以下の場合はシングルトークでない可能性が高く、前回のエコー消去量推定値b(k-1)に係数γを乗算した値をエコー消去量推定値b(k)として通話状態算出部30に与える。
Step S7: The smaller one of the predetermined value ε and the ratio r B is output as the call state value c (k) and given to the estimated value selecting unit 26 of the echo cancellation amount estimating unit 20. As described above, c (k) = ε is unconditionally set for the first time.
Step S8: It is determined whether the call state value c (k) is γ or less.
Step S9: If c (k) is less than or equal to γ, there is a high possibility that it is not single talk, and the value obtained by multiplying the previous echo cancellation amount estimated value b (k−1) by the coefficient γ is used as the echo cancellation amount estimated value b (k ) To the call state calculation unit 30.

ステップS10:c(k)がガンマ以下でない場合は、シングルトークの可能性が高いので、ステップS5で求めた暫定エコー消去量推定値b’(k)をエコー消去量推定値b(k)として通話状態値算出部30に与える。
ステップS11:k←k+1と歩進して次の時点(k+1)についてステップS2〜S10を実行する。
以上の処理を順次kの値をステップS11で歩進するごとに繰り返す。
図4のステップS7で得られた通話状態値c(k)、即ち図3の通話状態値算出部30により生成された通話状態値c(k)は、シングルトークである可能性の大きさを表しており、この値により、例えば係数更新部14でのタップ係数h’(k)の更新ステップサイズを制御することにより、c(k)が1より小さくシングルトークの可能性が小さい場合でもタップ係数の更新ステップサイズαに対し1より小さな値を乗算することで更新誤差が小さくなるようにできる。シングルトークの状態からシングルトークでない状態まで通話状態値c(k)は連続的に変化するので、タップ係数の更新ステップサイズを連続的に変化させることができる。そのため、エコー消去量が急激に劣化することを防ぐことができる。
Step S10: When c (k) is not less than or equal to gamma, the possibility of single talk is high, so the provisional echo cancellation amount estimated value b ′ (k) obtained in step S5 is used as the echo cancellation amount estimated value b (k). This is given to the call state value calculation unit 30.
Step S11: Steps k2 to k + 1 and steps S2 to S10 are executed for the next time point (k + 1).
The above processing is sequentially repeated every time the value k is incremented in step S11.
The call state value c (k) obtained in step S7 in FIG. 4, that is, the call state value c (k) generated by the call state value calculation unit 30 in FIG. 3, indicates the magnitude of the possibility of single talk. By this value, for example, by controlling the update step size of the tap coefficient h ′ (k) in the coefficient update unit 14, even when c (k) is smaller than 1 and the possibility of single talk is small, the tap is performed. The update error can be reduced by multiplying the coefficient update step size α by a value smaller than 1. Since the call state value c (k) continuously changes from the single talk state to the non-single talk state, the tap coefficient update step size can be continuously changed. For this reason, it is possible to prevent the echo cancellation amount from being rapidly deteriorated.

図5は通話状態値c(k)によりタップ係数h’(k)の更新ステップサイズを制御するように構成した係数更新部14の例を示す。この係数更新部14において、受話信号ベクトルx(k-1)はホールド部14Aに保持され、エコー消去信号e(k-1)はホールド部14Bに保持され、これら受話信号ベクトルx(k-1)とエコー消去信号e(k-1)が乗算器14Cにより乗算される。一方、受話信号ベクトルx(k-1)はベクトル乗算部14Dに与えられてxT(k-1)x(k-1)+δが計算される。割算部14Eはe(k-1)x(k-1)とxT(k-1)x(k-1)+δの比を求める。その比は乗算器14Fでステップサイズαと乗算され、その乗算結果にさらに乗算器14Gで通話状態値c(k)が乗算される。ホールド部14Jには前回のタップ係数h’(k-1)が保持されており、そのタップ係数h’(k-1)と乗算器14Gの出力とが加算部14Hで加算され、更新されたタップ係数h’(k)とされる。即ち、乗算器14Fの出力にさらに通話状態値c(k)を乗算することにより、式(1)におけるステップサイズαの代わりにαc(k)を使用することになる。
実施例2
前述の実施例では通話状態値c(k)を求めるためピークホールド値の比として式(5), (6)に示されるように対数パワーのピークホールド値の比を使用する例を示したが、この実施例2では、以下に示すように振幅絶対値のピークホールド値の比を用いる。即ち、通話状態値c(k)として式(5), (6)の代わりに次式(11), (12)を使用する。
FIG. 5 shows an example of the coefficient updating unit 14 configured to control the update step size of the tap coefficient h ′ (k) based on the call state value c (k). In this coefficient updating unit 14, the received signal vector x (k-1) is held in the hold unit 14A, and the echo cancellation signal e (k-1) is held in the hold unit 14B, and these received signal vectors x (k-1). ) And the echo cancellation signal e (k-1) are multiplied by the multiplier 14C. On the other hand, the received signal vector x (k−1) is given to the vector multiplier 14D to calculate x T (k−1) x (k−1) + δ. Dividing unit 14E obtains a ratio of e (k-1) x ( k-1) and x T (k-1) x (k-1) + δ. The ratio is multiplied by the step size α by the multiplier 14F, and the multiplication result is further multiplied by the call state value c (k) by the multiplier 14G. The hold unit 14J holds the previous tap coefficient h ′ (k−1), and the tap coefficient h ′ (k−1) and the output of the multiplier 14G are added and updated by the adder 14H. The tap coefficient is h ′ (k). That is, by multiplying the output of the multiplier 14F by the call state value c (k), αc (k) is used instead of the step size α in the equation (1).
Example 2
In the above-described embodiment, an example in which the ratio of the peak hold value of the logarithmic power is used as the ratio of the peak hold value to obtain the call state value c (k) as shown in the equations (5) and (6) is shown. In the second embodiment, the ratio of the peak hold value of the absolute amplitude value is used as shown below. That is, the following expressions (11) and (12) are used as the call state value c (k) instead of the expressions (5) and (6).

Figure 2006148540
ここで、|e(k-i)|+δ=E(k-i)および|y(k-i)|+δ=Y(k-i)とおけば、式(11), (12)中の信号e(k)およびy(k)に関するピークホールドPH{}は式(9), (10)で表される。式(11), (12)から明らかなように、ここではエコー消去信号e(k)および収音信号y(k)の振幅の絶対値のピークホールド値の比をi=0〜L-1に渡って積算している点が式(5), (6)と異なる。
図6は式(11), (12)に基づく通話状態値装置100の構成例を示す。この構成は、図3の構成におけるサンプルパワー算出部21E2, 21Y2の代わりに振幅絶対値算出部21E2’, 21Y2’が設けられており、加算器22Aの代わりに除算器22A’が設けられている構成となっている。振幅絶対値算出部21E2’はエコー消去信号e(k-i)の絶対値に微小値δを加算してE(k-i)=|e(k-i)|+δを求め、以下図3の場合と同様に乗算器21E3、ピーク値選択部21E4、ピークホールド部21E5により式(7), (9)に従ってピークホールド値PH{E(k-i)}を求める。
Figure 2006148540
Here, if | e (ki) | + δ = E (ki) and | y (ki) | + δ = Y (ki), the signals e (k) and y ( The peak hold PH {} relating to k) is expressed by equations (9) and (10). As is clear from the equations (11) and (12), here, the ratio of the peak hold value of the absolute value of the amplitude of the echo cancellation signal e (k) and the collected sound signal y (k) is i = 0 to L−1. Is different from Eqs. (5) and (6).
FIG. 6 shows a configuration example of the call state value device 100 based on the equations (11) and (12). In this configuration, amplitude absolute value calculation units 21E2 ′ and 21Y2 ′ are provided instead of the sample power calculation units 21E2 and 21Y2 in the configuration of FIG. 3, and a divider 22A ′ is provided instead of the adder 22A. It has a configuration. The amplitude absolute value calculation unit 21E2 ′ adds a small value δ to the absolute value of the echo cancellation signal e (ki) to obtain E (ki) = | e (ki) | + δ, and then performs multiplication similarly to the case of FIG. The peak hold value PH {E (ki)} is obtained according to the equations (7) and (9) by the device 21E3, the peak value selection unit 21E4, and the peak hold unit 21E5.

ピークホールド算出部21Yにおいても同様にして、収音信号y(k)の振幅絶対値Y(k-i)=|y(k-i)|+δのピークホールド値PH{Y(k-i)}を求める。このようにして得たピークホールド値の比PH{E(k-i)}/PH{Y(k-i)}を除算器22A’で算出し、得られたピークホールド比を総和部22B’でi=0, 1, …, L-1について順次積算し、その演算結果に乗算器23で1/Lを乗算して暫定エコー消去量推定値b’(k)を得る。以下の処理は図3の実施例と同様であり、説明を省略する。
なお、図3および6において、乗算器23は区間長Lでのピークホールド比の積算値を平均化するために1/Lを乗算(Lで除算)しているが、Lは予め決めた一定値なので、Lで除算しなくてもよい。また、Lの値は例えば1〜4096の範囲から実験により最適なものを決める。L=1の場合は式(5), (6), (11), (12)から明らかなように、エコー消去信号および収音信号の現時点のサンプルe(k), y(k)のピークホールド値から通話状態値c(k)を求めることになるが、その場合でもピークホールド値を使用するため従来のピークホールドを行わない場合と比べ、より信頼度の高い通話状態値c(k)が得られる。
実施例3
図7は本発明装置の第2の実施例をエコーキャンセラ10に適用して示す。この実施例の通話状態値算出装置100は、図2の実施例における通話状態値算出装置100において、さらに第2通話状態値算出部40を追加した構成となっている。第2通話状態値算出部40は、受話信号x(k)のレベルから受話信号の有無の判定を行い、また、収音信号y(k)のレベルカラム信号状態か否かの判定を行い、両者の関係から、通話状態として、シングルトーク状態、ダブルトーク状態、受話信号が検出されず送話信号が検出される状態、無信号状態、の4つの状態を判別し、その状態を表す第2通話状態値c’(k)を生成する。
Similarly, the peak hold calculation unit 21Y obtains the peak hold value PH {Y (ki)} of the amplitude absolute value Y (ki) = | y (ki) | + δ of the collected sound signal y (k). The peak hold value ratio PH {E (ki)} / PH {Y (ki)} thus obtained is calculated by the divider 22A ′, and the obtained peak hold ratio is calculated by i = 0 in the summation unit 22B ′. , 1,..., L−1 are sequentially integrated, and the operation result is multiplied by 1 / L by the multiplier 23 to obtain a provisional echo cancellation amount estimated value b ′ (k). The following processing is the same as that of the embodiment of FIG.
3 and 6, the multiplier 23 multiplies (divides by L) 1 / L in order to average the integrated value of the peak hold ratio in the section length L, where L is a predetermined constant. Since it is a value, it is not necessary to divide by L. The optimum value of L is determined by experiment from the range of 1 to 4096, for example. When L = 1, as is clear from the equations (5), (6), (11), (12), the peak of the current samples e (k), y (k) of the echo cancellation signal and the collected sound signal The call state value c (k) is calculated from the hold value. Even in this case, the peak hold value is used, so the call state value c (k) is more reliable than the conventional case where peak hold is not performed. Is obtained.
Example 3
FIG. 7 shows a second embodiment of the apparatus of the present invention applied to the echo canceller 10. The call state value calculation device 100 of this embodiment has a configuration in which a second call state value calculation unit 40 is further added to the call state value calculation device 100 of the embodiment of FIG. The second call state value calculation unit 40 determines the presence / absence of a received signal from the level of the received signal x (k), determines whether the sound pickup signal y (k) is in the level column signal state, From the relationship between the two, four states of a single talk state, a double talk state, a state where no incoming signal is detected and a transmitted signal is detected, and a no-signal state are discriminated as a call state, and a second state representing the state. A call state value c ′ (k) is generated.

ここで、無信号状態は収音信号y(k)のレベルが極めて小さい状態を指す。通話状態について、例えば、通話状態値c(k)が低く、第2通話状態値算出部40において受話信号有りと判定した場合はダブルトーク状態、受話信号無しと判定した場合は受話信号が検出されず送話信号が検出される状態と推定することができる。また、収音信号y(k)が無信号のとき無信号状態になる。このような通話状態を詳細に区別できる第2通話状態値c'(k)は、以下の式(13), (14)によって表される。

Figure 2006148540
Here, the no-signal state refers to a state where the level of the collected sound signal y (k) is extremely small. As for the call state, for example, when the call state value c (k) is low and the second call state value calculation unit 40 determines that there is a received signal, a double talk state is detected, and when it is determined that there is no received signal, a received signal is detected. Therefore, it can be estimated that a transmission signal is detected. Further, when the collected sound signal y (k) is no signal, the no-signal state is set. The second call state value c ′ (k) that can distinguish such a call state in detail is expressed by the following equations (13) and (14).
Figure 2006148540

式(14)のように収音信号y(k)の最新のLこのサンプルについて振幅の絶対値の平均を求め、その平均と予め決めた値κと比較し、κより小であれば無信号状態と判定し、フラグflg(k)を0に設定する。κは信号の有無を決定する係数である。κより小でない場合は送話信号のみが検出された状態と判定し、フラグflg(k)を1に設定する。その他の場合(otherwise)、即ちκより小でなければシングルトークかダブルトークの状態であり、フラグflg(k)を2に設定する。シングルトークかダブルトークの状態では、第1通話状態値c(k)が0に近ければダブルトークの可能性が大である。   The latest L of the collected sound signal y (k) as in the equation (14), the average of the absolute value of the sample is obtained, and the average is compared with a predetermined value κ. The state is determined, and the flag flg (k) is set to 0. κ is a coefficient that determines the presence or absence of a signal. If it is not smaller than κ, it is determined that only the transmission signal is detected, and the flag flg (k) is set to 1. In the other case, that is, if it is not smaller than κ, the state is single talk or double talk, and the flag flg (k) is set to 2. In the single talk or double talk state, if the first call state value c (k) is close to 0, the possibility of double talk is great.

これらの判定に基づいて、式(13)により、flg(k)=0の場合はc’(k)=-2に設定し、flg(k)=1であればc’(k)=-1に設定し、それ以外ではシングルトークまたはダブルトークの状態としてc’(k)=c(k)に設定し、第2通話状態値c’(k)を出力する。即ち、第2通話状態値c'(k)=-2のとき無信号状態、c’(k)=-1のとき受話信号が検出されず送話信号が検出される状態、c’(k)=c(k)のときシングルトーク状態もしくはダブルトーク状態を表す。
この実施例で生成された第2通話状態値c’(k)は、図7に示すように例えば受話信号x(k)が入力されるエコーキャンセラ10の入力側に設けた損失挿入部51およびエコー消去信号e(k)が送出されるエコーキャンセラ10の送出側に設けた損失挿入部52の挿入損失を制御するために使用することができる。即ち、損失挿入部51では、与えられた第2通話状態値c’(k)の大きさに応じて利得が大きくなるように挿入損失を制御する。損失挿入部52においては、c’(k)の大きさに応じて損失が大きくなるように挿入損失を制御する。
Based on these determinations, according to equation (13), when flg (k) = 0, c ′ (k) = − 2 is set, and when flg (k) = 1, c ′ (k) = − 1 is set, otherwise c ′ (k) = c (k) is set as a single talk or double talk state, and a second call state value c ′ (k) is output. That is, when the second call state value c ′ (k) = − 2, there is no signal state, when c ′ (k) = − 1, the state where the reception signal is not detected and the transmission signal is detected, c ′ (k ) = c (k) indicates a single talk state or a double talk state.
As shown in FIG. 7, the second call state value c ′ (k) generated in this embodiment includes, for example, a loss insertion unit 51 provided on the input side of the echo canceller 10 to which the received signal x (k) is input and This can be used to control the insertion loss of the loss insertion unit 52 provided on the transmission side of the echo canceller 10 to which the echo cancellation signal e (k) is transmitted. That is, the loss insertion unit 51 controls the insertion loss so that the gain increases according to the magnitude of the given second call state value c ′ (k). In the loss insertion unit 52, the insertion loss is controlled so that the loss increases according to the magnitude of c ′ (k).

図8は第2通話状態値算出部40の構成例を示す。受話信号x(k)の最新のL個のサンプルが記憶部41Xに保持され、それらL個のサンプルの振幅の絶対値を加算部42Xで加算する。加算結果は乗算器43Xで1/Lが乗算され、平均振幅Ax=Σ|x(k)|/Lを得る。判定部44Xは平均振幅Axを所定値κと比較し、判定結果Y/Nをフラグ設定部45に与える。動揺に、収音信号y(k)についても記憶部41Yに保持されたL個のサンプルを使って加算部42Y、乗算器43Yにより平均振幅Ay=Σ|y(k)|/Lを求め、判定部44Yでκ>Ayかを判定し、判定結果Y/Nをフラグ設定部45に与える。   FIG. 8 shows a configuration example of the second call state value calculation unit 40. The latest L samples of the received signal x (k) are stored in the storage unit 41X, and the absolute values of the amplitudes of the L samples are added by the adding unit 42X. The addition result is multiplied by 1 / L by the multiplier 43X to obtain an average amplitude Ax = Σ | x (k) | / L. The determination unit 44X compares the average amplitude Ax with a predetermined value κ, and gives the determination result Y / N to the flag setting unit 45. For the sound pickup signal y (k), the average amplitude Ay = Σ | y (k) | / L is obtained by the adder 42Y and the multiplier 43Y using the L samples stored in the storage unit 41Y. The determination unit 44Y determines whether κ> Ay, and gives the determination result Y / N to the flag setting unit 45.

フラグ設定部45は式(14)に従ってフラグflg(k)を設定し、第2通話状態生成部46に与える。第2通話状態値生成部46は式(13)に基づいて第2通話状態値c’(k)を生成し出力する。
このような第2通話状態値c’(k)による制御により、エコーを抑圧するとともに、ダブルトーク時の送話信号を抑圧することができる。しかも、シングルトークまたはダブルトーク時にはシングルトークの可能性の大きさに応じた値のc’(k)=c(k)が損失挿入部51,52に与えられるので、よりきめの細かい挿入損失の制御が可能となる。
The flag setting unit 45 sets the flag flg (k) according to the equation (14) and gives it to the second call state generation unit 46. The second call state value generation unit 46 generates and outputs a second call state value c ′ (k) based on the equation (13).
By such control using the second call state value c ′ (k), it is possible to suppress echoes and suppress transmission signals during double talk. In addition, since c ′ (k) = c (k) corresponding to the magnitude of the possibility of single talk is given to the loss insertion sections 51 and 52 during single talk or double talk, the finer insertion loss can be reduced. Control becomes possible.

以上説明したように、この発明によれば、通話状態値c(k)は式(5), (6)または式(11), (12)で示されるように、最新のLサンプルに渡ってピークホールド値の比を順次積算して求めるので、従来のような各離散時間の振幅比の誤差を避けることができる。さらに、通話状態値c(k)から通話状態がシングルトークであることを判定する場合、エコー消去信号e(k)と収音信号y(k)のピーク積算値の比から求めたエコー消去量b(k)を使用するので、ダブルトーク検出精度が入力信号に大きく依存するという従来の問題を解決できる。   As described above, according to the present invention, the call state value c (k) is obtained over the latest L samples as shown in the equations (5), (6) or (11), (12). Since the ratios of the peak hold values are sequentially accumulated, an error in the amplitude ratio at each discrete time as in the prior art can be avoided. Furthermore, when it is determined from the call state value c (k) that the call state is single talk, the amount of echo cancellation obtained from the ratio of the peak integrated value of the echo cancellation signal e (k) and the collected sound signal y (k) Since b (k) is used, the conventional problem that the double-talk detection accuracy greatly depends on the input signal can be solved.

回線エコーキャンセラ制御、ハンズフリー通話、ハンズフリー音声認識などへの適用が可能である。   It can be applied to line echo canceller control, hands-free calling, hands-free speech recognition, and the like.

従来のエコーキャンセラの基本構成を示すブロック図。The block diagram which shows the basic composition of the conventional echo canceller. 本発明の第1実施例による通話状態値算出装置100を適用したエコーキャンセラを示すブロック図。The block diagram which shows the echo canceller to which the call state value calculation apparatus 100 by 1st Example of this invention is applied. 通話状態値算出装置の構成例を示すブロック図。The block diagram which shows the structural example of a call state value calculation apparatus. 通話状態値算出の処理手順を示すフロー図。The flowchart which shows the process sequence of call state value calculation. 係数更新部の構成例を示すブロック図。The block diagram which shows the structural example of a coefficient update part. この発明の第2の実施例による通話状態値算出装置の構成例を示すブロック図。The block diagram which shows the structural example of the call state value calculation apparatus by 2nd Example of this invention. 第3実施例による通話状態値算出装置100を適用したエコーキャンセラを示すブロック図。The block diagram which shows the echo canceller to which the call state value calculation apparatus 100 by 3rd Example is applied. 第2通話状態値算出部40の構成例を示すブロック図。The block diagram which shows the structural example of the 2nd call state value calculation part.

Claims (5)

一連の離散的時点の時点kにおけるエコー消去信号とエコー消去前信号のそれぞれ一定サンプル数Lに渡るピークホールドの比の積算値と、与えられた通話状態値とからエコー消去量推定値を算出するエコー消去量推定部と、kは任意の整数であり、Lは1以上の予め決めた整数であり、
時点(k+1)におけるエコー消去信号とエコー消去前信号のそれぞれ上記一定サンプル数に渡るピークホールドの比の積算値と、上記エコー消去量推定値とから上記通話状態値を算出し、上記エコー消去量推定部に与える通話状態値算出部、
とを含むことを特徴とする通話状態値算出装置。
An echo cancellation amount estimated value is calculated from the integrated value of the ratio of peak hold over a fixed number of samples L of the echo cancellation signal and the pre-echo cancellation signal at a series of discrete time points k and the given call state value. The echo cancellation amount estimation unit, k is an arbitrary integer, L is a predetermined integer of 1 or more,
The call state value is calculated from the integrated value of the peak hold ratio over the fixed number of samples of the echo cancellation signal and the signal before echo cancellation at time (k + 1) and the echo cancellation amount estimated value, Call state value calculation unit to be given to the erasure amount estimation unit,
A call state value calculation device comprising:
請求項1記載の通話状態値算出装置において、上記エコー消去量推定部は、上記エコー消去信号と上記エコー消去前信号のパワーのピークホールド値をそれぞれ求める第1および第2ピークホールド算出部と、上記エコー消去信号とエコー消去前信号のピークホールド値の比を最新のL個のサンプル時点に渡って順次加算して上記比の積算値を求めるピーク比積算部とを含むことを特徴とする通話状態値算出装置。   The call state value calculation apparatus according to claim 1, wherein the echo cancellation amount estimation unit includes first and second peak hold calculation units that respectively determine peak hold values of power of the echo cancellation signal and the signal before echo cancellation, And a peak ratio integrating unit that sequentially adds the ratio of the peak hold value of the echo cancellation signal and the signal before echo cancellation over the latest L sampling points to obtain an integrated value of the ratio. State value calculation device. 請求項1記載の通話状態値算出装置において、上記エコー消去量推定部は、上記エコー消去信号と上記エコー消去前信号の振幅絶対値のピークホールド値をそれぞれ求める第1および第2ピークホールド算出部と、上記エコー消去信号および上記エコー消去前信号のピークホールド値の比を最新のL個のサンプル時点に渡って順次加算して上記比の積算値を求めるピーク比積算部とを含むことを特徴とする通話状態値算出装置。   2. The call state value calculation apparatus according to claim 1, wherein the echo cancellation amount estimation unit is a first peak hold calculation unit and a second peak hold calculation unit which respectively determine peak hold values of amplitude absolute values of the echo cancellation signal and the signal before echo cancellation. And a peak ratio integration unit that sequentially adds a ratio of peak hold values of the echo cancellation signal and the signal before echo cancellation over the latest L sampling points to obtain an integrated value of the ratio. Call state value calculation device. 請求項1、2または3のいずれか記載の通話状態値算出装置において、さらに、受話信号とエコー消去前信号のそれぞれの一定時間長の平均と、上記通話状態値から通信状態として無信号状態、シングルトーク状態、ダブルトーク状態、送話信号のみの状態を判別し、それらの状態に対応した第2の通話状態値を生成する第2通話状態値算出部が設けられていることを特徴とする通話状態値算出装置。   In the call state value calculation device according to any one of claims 1, 2, and 3, further, an average of each fixed time length of the received signal and the signal before echo cancellation, and a no signal state as a communication state from the call state value, A second call state value calculation unit is provided that determines a single talk state, a double talk state, and a state of only a transmission signal and generates a second call state value corresponding to those states. Call state value calculation device. 一連の離散的時点の時点kにおけるエコー消去信号とエコー消去前信号のそれぞれ一定サンプル数Lに渡るピークホールドの比の積算値を求める工程と、kは任意の整数であり、Lは1以上の任意の整数であり、
上記比の積算値と上記与えられた通話状態値とからエコー消去量推定値を算出する工程と、
時点(k+1)におけるエコー消去信号とエコー消去前信号のそれぞれ一定サンプル数Lに渡るピークホールドの比の積算値と、上記エコー消去量推定値とから通話状態値を算出する工程、
とを含むことを特徴とする通話状態値算出方法。
A step of obtaining an integrated value of peak hold ratios over a fixed number of samples L of the echo cancellation signal and the pre-echo cancellation signal at a series of discrete time points k, k is an arbitrary integer, and L is 1 or more Any integer,
Calculating an echo cancellation amount estimated value from the integrated value of the ratio and the given call state value;
Calculating a call state value from the integrated value of the ratio of peak hold over a fixed number of samples L of the echo cancellation signal and the signal before echo cancellation at time (k + 1) and the echo cancellation amount estimated value;
The call state value calculation method characterized by including.
JP2004335900A 2004-11-19 2004-11-19 Call state value calculation device and calculation method Expired - Fee Related JP4452162B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2004335900A JP4452162B2 (en) 2004-11-19 2004-11-19 Call state value calculation device and calculation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2004335900A JP4452162B2 (en) 2004-11-19 2004-11-19 Call state value calculation device and calculation method

Publications (2)

Publication Number Publication Date
JP2006148540A true JP2006148540A (en) 2006-06-08
JP4452162B2 JP4452162B2 (en) 2010-04-21

Family

ID=36627710

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2004335900A Expired - Fee Related JP4452162B2 (en) 2004-11-19 2004-11-19 Call state value calculation device and calculation method

Country Status (1)

Country Link
JP (1) JP4452162B2 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2013005106A (en) * 2011-06-14 2013-01-07 Nippon Telegr & Teleph Corp <Ntt> In-house sound amplification system, in-house sound amplification method, and program therefor
CN103259563A (en) * 2012-02-16 2013-08-21 联芯科技有限公司 Self-adapting filter divergence detection method and echo cancellation system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2013005106A (en) * 2011-06-14 2013-01-07 Nippon Telegr & Teleph Corp <Ntt> In-house sound amplification system, in-house sound amplification method, and program therefor
CN103259563A (en) * 2012-02-16 2013-08-21 联芯科技有限公司 Self-adapting filter divergence detection method and echo cancellation system

Also Published As

Publication number Publication date
JP4452162B2 (en) 2010-04-21

Similar Documents

Publication Publication Date Title
JP4697465B2 (en) Signal processing method, signal processing apparatus, and signal processing program
JP2538176B2 (en) Eco-control device
US10244121B2 (en) Automatic tuning of a gain controller
EP2330752B1 (en) Echo cancelling device
EP1714398B1 (en) Echo canceller with interference-level controlled step size
JPH114288A (en) Echo canceler device
WO2005125168A1 (en) Echo canceling apparatus, telephone set using the same, and echo canceling method
JP5430990B2 (en) Signal processing method, apparatus and program
JP4457639B2 (en) Echo canceller
JP2856113B2 (en) Echo canceller
EP2380339B1 (en) Determining an acoustic coupling between a far-end talker signal and a combined signal
WO2001054296A1 (en) Sound communication device and echo processor
US6185299B1 (en) Adaptive echo cancellation device in a voice communication system
JP5293952B2 (en) Signal processing method, signal processing apparatus, and signal processing program
EP0731593A2 (en) Echo removing apparatus
JP4452162B2 (en) Call state value calculation device and calculation method
JP5057109B2 (en) Echo canceller
JP2006279191A (en) Echo canceller
JP3244416B2 (en) Echo canceller
JP2650208B2 (en) Echo canceller
EP0715407B1 (en) Method and apparatus for controlling coefficients of adaptive filter
KR100431965B1 (en) Apparatus and method for removing echo-audio signal using time-varying algorithm with time-varying step size
JPH1168518A (en) Coefficient updating circuit
JPS63238727A (en) Method for controlling double talk detection of echo canceler
JP3635644B2 (en) Echo canceller

Legal Events

Date Code Title Description
RD03 Notification of appointment of power of attorney

Free format text: JAPANESE INTERMEDIATE CODE: A7423

Effective date: 20061225

A621 Written request for application examination

Free format text: JAPANESE INTERMEDIATE CODE: A621

Effective date: 20070126

A131 Notification of reasons for refusal

Free format text: JAPANESE INTERMEDIATE CODE: A131

Effective date: 20091104

A521 Written amendment

Free format text: JAPANESE INTERMEDIATE CODE: A523

Effective date: 20091218

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: 20100119

A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

A61 First payment of annual fees (during grant procedure)

Free format text: JAPANESE INTERMEDIATE CODE: A61

Effective date: 20100129

R150 Certificate of patent or registration of utility model

Ref document number: 4452162

Country of ref document: JP

Free format text: JAPANESE INTERMEDIATE CODE: R150

Free format text: JAPANESE INTERMEDIATE CODE: R150

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

Free format text: PAYMENT UNTIL: 20130205

Year of fee payment: 3

S531 Written request for registration of change of domicile

Free format text: JAPANESE INTERMEDIATE CODE: R313531

R350 Written notification of registration of transfer

Free format text: JAPANESE INTERMEDIATE CODE: R350

LAPS Cancellation because of no payment of annual fees