JP2002064617A - Echo suppression method and echo suppression equipment - Google Patents
Echo suppression method and echo suppression equipmentInfo
- Publication number
- JP2002064617A JP2002064617A JP2000246192A JP2000246192A JP2002064617A JP 2002064617 A JP2002064617 A JP 2002064617A JP 2000246192 A JP2000246192 A JP 2000246192A JP 2000246192 A JP2000246192 A JP 2000246192A JP 2002064617 A JP2002064617 A JP 2002064617A
- Authority
- JP
- Japan
- Prior art keywords
- signal
- filter
- reverberation
- echo
- echo suppression
- 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
Links
Landscapes
- Circuit For Audible Band Transducer (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
- Telephone Function (AREA)
Abstract
Description
【０００１】[0001]
【発明の属する技術分野】この発明は、拡声通話系にお
いてハウリングの原因となる反響を抑圧する技術に関す
る。[0001] 1. Field of the Invention [0002] The present invention relates to a technique for suppressing reverberation which causes howling in a loudspeaker system.
【０００２】[0002]
【従来の技術】拡声通話系では、スピーカから拡声され
た受話音声がマイクに収音されて生じる反響が問題とな
る。この反響現象は音響エコーとも呼ばれ、対地の拡声
通話系を含めて形成される閉ループのループゲインが１
より大きい場合にはハウリングとなり、通話が不可能と
なる。またループゲインが１より小さい場合にも通話の
障害や不快感などの悪影響を生じる。2. Description of the Related Art In a loudspeaker system, there is a problem of reverberation generated when a received voice loudspeaked from a speaker is picked up by a microphone. This reverberation phenomenon is also called an acoustic echo, and the loop gain of a closed loop formed including a loudspeaker system over the ground is 1
If it is larger, howling will occur and the call will not be possible. Also, when the loop gain is smaller than 1, adverse effects such as trouble in communication and discomfort occur.
【０００３】このような拡声通話系の問題点を解決する
方法として、受話信号から反響を模擬し、収音信号から
模擬反響信号を差し引いた残差信号を送信信号とする反
響消去法がある。しかし通常反響は完全に消去されない
ため、送話検出法により送話の有無を調べ、送話のない
ときに残留エコーを聴感上検知できないレベルまで減衰
させる反響抑圧方法と組み合わせて用いられることが多
い。この反響抑圧方法としては全帯域の振幅を一律に減
衰させる処理が行われるが、エコーと同時に背景音や送
話に検出されなかった送話音声も一緒に抑圧する。この
ため、実際の拡声通信においては抑圧処理のオンオフが
断続感を与え、通信品質を低いと感じさせてしまう問題
がある。As a method of solving such a problem of the loudspeaker communication system, there is an echo cancellation method in which reverberation is simulated from a reception signal, and a residual signal obtained by subtracting a simulated reverberation signal from a collected signal is used as a transmission signal. However, since the echo is usually not completely eliminated, it is often used in combination with an echo suppression method that checks for the presence or absence of a speech using a speech detection method and attenuates the residual echo to a level that cannot be detected audibly when there is no speech. . As the echo suppression method, a process of uniformly attenuating the amplitude of the entire band is performed, but simultaneously with the echo, the background sound and the transmitted voice not detected in the transmitted voice are also suppressed. For this reason, there is a problem that in the actual loudspeaker communication, the on / off of the suppression processing gives an intermittent feeling, and the communication quality is felt to be low.
【０００４】この問題を解決するため、受話信号と残差
信号の短時間スペクトルから周波数成分毎に反響抑圧ゲ
インを決定して収音信号の短時間スペクトルにこのゲイ
ンを乗算することで、反響を抑圧して送信信号とする方
法がすでに提案されている。この方法は、残留音響エコ
ーを充分減衰させるものの背景音はほとんど減衰されな
いため、良好の通話品質が確保される。In order to solve this problem, the reverberation suppression gain is determined for each frequency component from the short-time spectrum of the received signal and the residual signal, and the short-time spectrum of the picked-up signal is multiplied by this gain. A method of suppressing a transmission signal has already been proposed. In this method, the residual sound echo is sufficiently attenuated, but the background sound is hardly attenuated, so that good speech quality is ensured.
【０００５】[0005]
【発明が解決しようとする課題】この反響抑圧方法は、
一定長の時間フレーム毎に受話信号と残留反響信号の短
時間スペクトルをＦＦＴ（高速フーリエ交換）により求
め、この情報から周波数成分毎に反響抑圧ゲインを求め
る。このゲインを周波数領域で残留反響信号の短時間ス
ペクトルに乗算した後に、逆ＦＦＴを用いて実際の時間
領域の信号を得ている。このため、入力された残留反響
信号が処理されて時間領域の信号として出力されるまで
の処理遅延は、ＦＦＴ及び逆ＦＦＴ処理に要する時間、
スペクトル処理に要する時間とフレーム長に相当する時
間の和になる。例えばサンプリング周波数が８ＫＨｚで
時間フレームが５１２サンプルからなるとき、この処理
遅延は少なくとも６４ｍｓ以上になる。SUMMARY OF THE INVENTION
A short-time spectrum of the received signal and the residual echo signal is obtained by FFT (fast Fourier exchange) for each fixed-length time frame, and an echo suppression gain is obtained for each frequency component from this information. After multiplying the short-time spectrum of the residual echo signal in the frequency domain by this gain, an actual time-domain signal is obtained using inverse FFT. For this reason, the processing delay until the input residual echo signal is processed and output as a signal in the time domain is the time required for the FFT and inverse FFT processing,
This is the sum of the time required for the spectrum processing and the time corresponding to the frame length. For example, when the sampling frequency is 8 KHz and the time frame consists of 512 samples, this processing delay is at least 64 ms or more.
【０００６】[0006]
【課題を解決するための手段】この発明では、ＦＦＴに
より受話信号の短時間スペクトルを求める代わりに、線
形予測法を用いて受話信号のスペクトル包絡を求める。
そして、このスペクトル包絡でパワーの大きい部分を抑
圧するＦＩＲフィルタの係数を線形予測係数から直接求
め、時間領域にてこのフィルタで収音信号を処理する。
この発明による反響抑圧方法によればＦＦＴを経由しな
いので、従来法と比較してほとんど処理遅延なしで実行
することが可能である。線形予測分析では、音声の発生
システムはAccording to the present invention, the spectrum envelope of the received signal is obtained by using a linear prediction method instead of obtaining the short-time spectrum of the received signal by FFT.
Then, a coefficient of an FIR filter for suppressing a high power portion in the spectrum envelope is directly obtained from the linear prediction coefficient, and the sound pickup signal is processed by the filter in a time domain.
According to the echo suppression method according to the present invention, since it does not pass through the FFT, it can be executed with almost no processing delay as compared with the conventional method. In linear predictive analysis, the speech generation system
【０００７】[0007]
【数１】 (Equation 1)
【０００８】というＡＲモデル（全極型デジタルフィル
タ）に、ピッチ信号に相当する１個のインパルスもしく
は数学的に等価な白色ランダム雑音が入力されるシステ
ムとして、モデル化される。次数ｐのＡＲモデルは、信
号ｘ（ｎ）と過去ｐ時点のｘ（ｎ）の値からの線形予測
値ｕ（ｎ）との誤差ｅ（ｎ）An AR model (all-pole digital filter) is modeled as a system in which one impulse corresponding to a pitch signal or mathematically equivalent white random noise is input. The AR model of order p is the error e (n) between the signal x (n) and the linear prediction value u (n) from the value of x (n) at the time point p in the past.
【０００９】[0009]
【数２】 (Equation 2)
【００１０】[0010]
【数３】 (Equation 3)
【００１１】を最小にする線形予測係数ａiとして求め
られる。ＡＲモデルの全極型デジタルフィルタ１／Ａ
（ｚ）は、ホルマントすなわち音声発生時の声道の共振
に対応し音声スペクトルの概形を与えることが知られて
いる。なお音声の線形予測分析については、例えば中田
和夫著、森北出版社発行『音声の高能率符号化』に記載
されている。残響時間が長くないとき、収音された反響
信号ｙ（ｎ）のスペクトル概形は受話信号ｘ（ｎ）のス
ペクトル外形とほぼ一致している。したがって、収音信
号を全帯域一律に抑圧する変わりに、受話信号パワーが
大きい周波数帯域で、収音信号を大きく抑圧し、受話信
号パワーの小さい周波数帯域で収音信号を小さく抑圧す
る信号処理を行うことで、背景音を通しつつ、反響を効
果的に抑圧することができる。Is obtained as a linear prediction coefficient a i that minimizes AR model all-pole digital filter 1 / A
It is known that (z) corresponds to the formant, that is, the resonance of the vocal tract at the time of voice generation, and gives an approximate shape of the voice spectrum. The linear prediction analysis of speech is described, for example, in Kazuo Nakata, "High-efficiency speech encoding", published by Morikita Shuppansha. When the reverberation time is not long, the spectral outline of the collected echo signal y (n) substantially matches the spectral outline of the received signal x (n). Therefore, instead of uniformly suppressing the sound pickup signal in the entire band, a signal processing for suppressing the sound pickup signal largely in the frequency band where the received signal power is large and suppressing the sound pickup signal small in the frequency band where the received signal power is small is used. By doing so, the reverberation can be effectively suppressed while passing the background sound.
【００１２】受話信号に線形予測法を適用し線形予測係
数が得られているとき、デジタルＦＩＲフィルタＡ
（ｚ）は、このような周波数特性を持っている。実際に
使用するフィルタとしては、例えばフィルタ周波数特性
の最大振幅が１になるようにＡ（ｚ）のゲインを調整し
たフィルタＢ（ｚ）等が考えられる。When a linear prediction method is applied to a received signal to obtain a linear prediction coefficient, a digital FIR filter A
(Z) has such a frequency characteristic. As a filter to be actually used, for example, a filter B (z) in which the gain of A (z) is adjusted so that the maximum amplitude of the filter frequency characteristic becomes 1 can be considered.
【００１３】[0013]
【発明の実施の形態】図１に請求項１に記載の反響抑圧
方法を適用した請求項６記載の反響抑圧装置の実施例を
示す。受話信号Ａ１は、再生器２により音響信号に再生
されると同時に、送信側減衰部１を構成するスペクトル
包絡推定部１１に入力され、このスペクトル包絡推定部
１１で線形予測分析により予測係数の形でスペクトル包
絡情報が抽出される。反響抑制フィルタ算出部１２は、
反響抑圧フィルタ１３が受話信号のスペクトルを平坦化
するフィルタ係数を予測係数から算出し、そのフィルタ
係数を反響抑圧フィルタ１３に設定する。収音器３から
の収音信号は、反響抑圧フィルタ１３による反響抑圧処
理を経て送信信号Ａ２として送信される。FIG. 1 shows an embodiment of a reverberation suppressing apparatus according to claim 6 to which the reverberation suppressing method according to claim 1 is applied. The received signal A1 is reproduced into an acoustic signal by the regenerator 2, and at the same time, is input to a spectrum envelope estimating unit 11 constituting the transmission-side attenuating unit 1, and the spectrum envelope estimating unit 11 forms a prediction coefficient by linear prediction analysis. Extracts the spectrum envelope information. The echo suppression filter calculation unit 12 calculates
The echo suppression filter 13 calculates a filter coefficient for flattening the spectrum of the received signal from the prediction coefficient, and sets the filter coefficient in the echo suppression filter 13. The sound pickup signal from the sound pickup device 3 is transmitted as a transmission signal A2 through echo suppression processing by the echo suppression filter 13.
【００１４】図２に請求項２に記載の反響抑圧方法を適
用した請求項７で提案する反響抑圧装置の一実施例を示
す。受話信号Ａ１は反響消去部４に入力されると同時
に、送信側減衰部１を構成するスペクトル包絡推定部１
１に入力され、このスペクトル包絡推定部１１で線形予
測分析により予測係数もしくは全極型フィルタの係数の
形でスペクトル包絡情報が抽出される。反響抑圧フィル
タ算出部１２は、請求項４に記載のように、この全極型
フィルタの分母部分を取出、フィルタの最大周波数維持
特性がほぼ１になるようにゲインを調整してＦＩＲフィ
ルタ係数を求め、このフィルタ係数を反響抑圧フィルタ
１３に設定する。FIG. 2 shows an embodiment of the reverberation suppressing apparatus proposed in claim 7 to which the reverberation suppressing method according to claim 2 is applied. The received speech signal A1 is input to the echo canceller 4 and at the same time, the spectrum envelope estimator 1
The spectrum envelope estimating unit 11 extracts spectrum envelope information in the form of prediction coefficients or coefficients of an all-pole filter by linear prediction analysis. The echo suppression filter calculation unit 12 extracts the denominator portion of the all-pole filter and adjusts the gain so that the maximum frequency maintaining characteristic of the filter becomes approximately 1, as described in claim 4, and adjusts the FIR filter coefficient. Then, this filter coefficient is set in the echo suppression filter 13.
【００１５】このようなＦＩＲフィルタの一例として、
線形予測の次数がｐで線形予測係数がａ1〜ａpのときにAs an example of such an FIR filter,
When the order of linear prediction is p and the linear prediction coefficients are a 1 to a p
【数４】 (Equation 4)
【００１６】というＦＩＲフィルタが考えられる。反響
消去部４に入力された受話信号は、再生器２により音響
信号に再生されると同時に擬似反響生成部４１に入力さ
れる。収音器３からの収音信号は、減算器４２で擬似反
響信号を差し引いた後に、反響抑圧フィルタ１３により
抑圧処理されて送信信号Ａ２となる。図３に請求項３に
記載の反響抑圧方法を適用した請求項８で提案する反響
抑圧装置の実施例を示す。音声が線形予測分析により高
能率で符号化・復号化して伝送される場合には、受信符
号ＡＡ１にすでにスペクトル包絡情報が含まれている。
抑圧フィルタ算出部１２は、復号化部１４により復号さ
れる前の受信符号ＡＡ１から直接反響抑圧フィルタの係
数（スペクトル包絡情報）を算出し、反響抑圧フィルタ
１３に設定することができる。送話信号は、反響抑圧フ
ィルタ１３により処理された後に符号化部１５により符
号化されて送信符号ＡＡ２として伝送される。この実施
例によれば線形予測分析の必要がないので、この処理分
の演算量が削減される。An FIR filter is conceivable. The received speech signal input to the echo canceller 4 is reproduced into an acoustic signal by the reproducer 2 and simultaneously input to the pseudo echo generator 41. The sound pickup signal from the sound pickup device 3 is subjected to a suppression process by the echo suppression filter 13 after subtracting the pseudo echo signal by the subtracter 42, and becomes a transmission signal A2. FIG. 3 shows an embodiment of the reverberation suppression device proposed in claim 8 to which the reverberation suppression method described in claim 3 is applied. In the case where speech is encoded and decoded with high efficiency by linear prediction analysis and transmitted, the received code AA1 already contains spectral envelope information.
The suppression filter calculation unit 12 can directly calculate the coefficient (spectrum envelope information) of the echo suppression filter from the received code AA1 before being decoded by the decoding unit 14, and can set the coefficient in the echo suppression filter 13. The transmission signal is processed by the echo suppression filter 13 and then encoded by the encoding unit 15 and transmitted as the transmission code AA2. According to this embodiment, since there is no need for linear prediction analysis, the amount of calculation for this processing is reduced.
【００１７】図４に請求項５に記載の反響抑圧方法を適
用した請求項１０で提案する反響抑圧装置の実施例を示
す。受話信号Ａ１は、再生器２により音響信号に再生さ
れると同時に、送信側減衰部１を構成する受話信号周波
数帯域分割部５に入力される。受話信号周波数帯域分割
部５では、入力信号は帯域通過フィルタによりＮ個の帯
域に分割され、予め決められた間引き率で間引かれて、
Ｎ個のサブバンド受話信号が得られる。収音器３からの
収音信号は、受話信号周波数帯域分割部５と同様の構成
を持つ収音信号周波数帯域分割部６に入力されＮ個のサ
ブバンド収音信号となる。Ｎ個の周波数帯域毎に、サブ
バンド受話信号とサブバンド収音信号がサブバンド反響
処理部７1〜７Nに入力され、周波数帯域毎にＮ個の反響
成分の減衰したサブバンド送信信号が得られる。送信信
号周波数帯域合成部８では、Ｎ個のサブバンド送信信号
から、送信信号を合成する。FIG. 4 shows an embodiment of a reverberation suppression apparatus proposed in claim 10 to which the reverberation suppression method described in claim 5 is applied. The reception signal A1 is reproduced by the reproducer 2 into an acoustic signal, and at the same time, is input to the reception signal frequency band division unit 5 constituting the transmission-side attenuation unit 1. In the reception signal frequency band division unit 5, the input signal is divided into N bands by a band-pass filter, and is thinned out at a predetermined thinning rate.
N subband reception signals are obtained. The sound pickup signal from the sound pickup device 3 is input to a sound pickup signal frequency band division unit 6 having the same configuration as the reception signal frequency band division unit 5, and becomes N subband sound pickup signals. Every N frequency bands, subband received signal subband collected sound signal is input to the sub-band echo processor 7 1 to 7-N, it is attenuated sub-band transmit signal of N reverberant component for each frequency band can get. The transmission signal frequency band synthesizing unit 8 synthesizes a transmission signal from the N subband transmission signals.
【００１８】一例として、図５のように０〜８ｋＨｚの
帯域を０〜４ｋＨｚと４〜８ｋＨｚの２帯域に分割し、
０〜４ｋＨｚ帯域では図２に示した反響消去部４として
動作する適応フィルタにより反響を消去し、４〜８ｋＨ
ｚ帯域では適応フィルタによる反響消去の代わりに請求
項１に記載の反響抑圧法を用いると、演算量を抑えつつ
広い帯域で、反響を効果的に抑えることが可能となる。
例えば通常の残響時間３００ｍｓ程度の会議室では、４
〜８ｋＨｚの帯域に適応フィルタを適用すると１０００
タップ（加算処理の段数）程度の畳込みを毎サンプル行
う必要がある。一方、請求項１に記載の反響抑圧方法を
適用すると、この発明による反響抑圧方法では受話信号
のスペクトル包絡のみに注目しているから、演算量は適
応フィルタの場合と比較して少なくてよく、毎サンプル
行う畳込みは、たかだか２０タップ程度で十分である。
また４ｋＨｚ以上の音声成分は一般にパワーが小さいた
めに、４〜８ｋＨｚの帯域の適応フィルタによる反響消
去は不十分になりがちであるが、請求項１に記載の反響
抑圧方法によれば確実に反響を抑圧しハウリングを阻止
することができる。As an example, a band of 0 to 8 kHz is divided into two bands of 0 to 4 kHz and 4 to 8 kHz as shown in FIG.
In the 0 to 4 kHz band, the echo is canceled by the adaptive filter operating as the echo canceller 4 shown in FIG.
In the z band, when the echo suppression method according to claim 1 is used instead of the echo cancellation by the adaptive filter, it is possible to effectively suppress the echo in a wide band while suppressing the amount of calculation.
For example, in a conference room with a normal reverberation time of about 300 ms, 4
When the adaptive filter is applied to the band of ~ 8 kHz, 1000
It is necessary to perform convolution of approximately taps (the number of stages of addition processing) for each sample. On the other hand, when the reverberation suppression method according to claim 1 is applied, the reverberation suppression method according to the present invention focuses only on the spectral envelope of the received signal, so that the amount of calculation may be smaller than that of the adaptive filter. For the convolution performed for each sample, at most about 20 taps are sufficient.
In general, since the power of voice components of 4 kHz or more generally has a small power, echo cancellation by an adaptive filter in a band of 4 to 8 kHz tends to be insufficient. However, the echo suppression method according to claim 1 surely provides an echo. And howling can be suppressed.
【００１９】[0019]
【効果の実証例】請求項２に記載の反響抑圧方法の性能
を調べるために、全帯域一律に振幅を減衰させる従来の
反響抑圧方法との比較を数値シミュレーションにより行
った。入力音声のサンプリング周波数は８ｋＨｚであ
り、線形予測分析の次数を１０次に、分析フレーム長を
１６ｍｓに設定した。また再生器２と収音器３の間の音
響結合量を５ｄＢとし、反響消去による反響の減衰量を
−１０ｄＢに設定した。[Examples of Demonstration of Effect] In order to investigate the performance of the echo suppression method according to the second aspect, a comparison with a conventional echo suppression method in which the amplitude is attenuated uniformly in all bands was performed by numerical simulation. The sampling frequency of the input speech was 8 kHz, the order of the linear prediction analysis was set to 10, and the analysis frame length was set to 16 ms. The acoustic coupling amount between the reproducing device 2 and the sound pickup device 3 was set to 5 dB, and the attenuation amount of the echo due to echo cancellation was set to -10 dB.
【００２０】図６に示す波形図において、図６Ａは反響
消去部通過後の残留反響と背景音の混在した信号、図６
Ｂは背景音、図６Ｃは全帯域を一律に抑圧する従来法に
よる送信信号、図６Ｄはこの発明の方法による送信信号
を示す。図７のグラフは、図６に示した各信号のスペク
トルグラムになっている。従来法による抑圧量は残留反
響のレベルがこの発明の方法と同じになるように−１８
ｄＢに設定した。In the waveform diagram shown in FIG. 6, FIG. 6A is a signal in which a residual echo and a background sound are mixed after passing through the echo canceller.
B is a background sound, FIG. 6C is a transmission signal according to the conventional method for uniformly suppressing the entire band, and FIG. 6D is a transmission signal according to the method of the present invention. The graph of FIG. 7 is a spectrumgram of each signal shown in FIG. The amount of suppression by the conventional method is -18 such that the level of the residual echo is the same as that of the method of the present invention.
It was set to dB.
【００２１】従来法では、残留反響と送話信号とが同時
に抑圧されてしまうために、背景音が殆ど送信されな
い。これに対し、この発明の方法では振幅が小さくなり
波形の形状が変わるものの背景音は送信され、反響は抑
圧されていることが分かる。図７に示したスペクトルグ
ラムからも、この発明の方法により反響がよく抑圧され
るものの背景音のスペクトルは残っていることがわか
る。In the conventional method, the background sound is hardly transmitted because the residual echo and the transmission signal are suppressed at the same time. On the other hand, according to the method of the present invention, although the amplitude is reduced and the shape of the waveform is changed, the background sound is transmitted and the reverberation is suppressed. It can also be seen from the spectrumgram shown in FIG. 7 that the reverberation is well suppressed by the method of the present invention, but the spectrum of the background sound remains.
【００２２】[0022]
【発明の効果】以上説明したように、この発明によれば
受話信号のスペクトル包絡を求めて反響抑圧フィルタ１
３のフィルタ係数を算出する方法を採ったから、ＦＦＴ
を用いる場合より処理遅延量を大幅に小さくすることが
できる利点が得られる。更に、全帯域を一律に抑圧して
反響を抑圧する方法と比較した場合、図６及び図７から
も明らかなように背景音の減衰を小さくすることができ
る。よって通話品質の劣化を抑えることができる利点も
得られる。As described above, according to the present invention, the echo suppression filter 1 determines the spectrum envelope of the received signal.
Since the method of calculating the filter coefficient of No. 3 was adopted, the FFT
Is advantageous in that the processing delay amount can be significantly reduced as compared with the case of using. Further, when compared with the method of uniformly suppressing the entire band to suppress reverberation, the attenuation of the background sound can be reduced as is clear from FIGS. 6 and 7. Therefore, the advantage that the deterioration of the communication quality can be suppressed can be obtained.
【００２３】また、この発明によれば話者音声と受話音
声が同時に存在するダブルトーク状態でも、受話音声の
線形予測分析に基づく反響抑圧フィルタを用いて反響の
周波数成分の通過を阻止する。背景音や話者音声がある
とき、適応フィルタが反響路の推定にもちいる参照信号
には受話音声の反響音とともにそれ以外の音も含まれる
ので、推定精度は背景音等のレベルにより制限されてし
まう。そのため、いかに高速に推定する適応フィルタを
用いても反響を消去しきれず、背景音等と同等レベルの
残留反響信号が背景音と共に送信されてしまう。一方、
本発明は受話音声情報から反響抑圧フィルタの係数を算
出し、背景音等に影響されない。そのため本発明を適応
フィルタと併用すると、残留反響の周波数成分を背景音
や話者音声のレベルから大幅に抑圧しつつ背景音や話者
音声の周波数成分を送信することが可能となる。Further, according to the present invention, even in the double talk state in which the speaker's voice and the received voice are present at the same time, the passage of the reverberation frequency component is prevented using the echo suppression filter based on the linear prediction analysis of the received voice. When there is background sound or speaker sound, the reference signal used by the adaptive filter for estimating the reverberation path includes other sounds as well as the reverberation sound of the received sound, so the estimation accuracy is limited by the level of the background sound and the like. Would. Therefore, no matter how fast the adaptive filter is used for estimation, the echo cannot be completely eliminated, and a residual echo signal having the same level as the background sound is transmitted together with the background sound. on the other hand,
The present invention calculates the coefficient of the echo suppression filter from the received voice information, and is not affected by background sound or the like. Therefore, when the present invention is used in combination with the adaptive filter, it becomes possible to transmit the frequency components of the background sound and the speaker voice while significantly suppressing the frequency components of the residual echo from the level of the background sound and the speaker voice.
【図面の簡単な説明】[Brief description of the drawings]
【図１】この発明の請求項１で提案する反響抑圧方法を
適用した請求項６で提案する反響抑圧装置の一実施例を
示すブロック図。FIG. 1 is a block diagram showing an embodiment of an echo suppression apparatus proposed in claim 6 to which the echo suppression method proposed in claim 1 of the present invention is applied.
【図２】この発明の請求項２で提案する反響抑圧方法を
適用した請求項７で提案する反響抑圧装置の一実施例を
示すブロック図。FIG. 2 is a block diagram showing an embodiment of an echo suppression apparatus proposed in claim 7 to which the echo suppression method proposed in claim 2 of the present invention is applied;
【図３】この発明の請求項３で提案する反響抑圧方法を
適用した請求項８で提案する反響抑圧装置の一実施例を
示すブロック図。FIG. 3 is a block diagram showing an embodiment of an echo suppression apparatus proposed in claim 8 to which the echo suppression method proposed in claim 3 of the present invention is applied.
【図４】この発明の請求項５で提案する反響抑圧方法を
適用した請求項１０で提案した反響抑圧装置の一実施例
を示すブロック図。FIG. 4 is a block diagram showing an embodiment of an echo suppression apparatus proposed in claim 10 to which the echo suppression method proposed in claim 5 of the present invention is applied;
【図５】図４に示した実施例の具体的な例を説明するた
めのブロック図。FIG. 5 is a block diagram for explaining a specific example of the embodiment shown in FIG. 4;
【図６】この発明の作用効果を説明するための波形図。FIG. 6 is a waveform chart for explaining the operation and effect of the present invention.
【図７】図６と同様のグラフ。FIG. 7 is a graph similar to FIG. 6;
１ 送信側減衰部 ２ 再生器 ３ 収音器 ４ 反響消去部 ５ 受話信号周波数帯域分割部 ６ 収音信号周波数帯域分割部 ７1〜７N サブバンド反響処理部 ８ 送信信号周波数帯域合成部 １１ スペクトル包絡推定部 １２ 反響抑圧フィルタ算出部 １３ 反響抑圧フィルタ １４ 復号化部 １５ 符号化部 Ａ１ 受話信号 Ａ２ 送信信号 ＡＡ１ 受信符号 ＡＡ２ 送信符号 ４１ 擬似反響生成部 ４２ 減算器 REFERENCE SIGNS LIST 1 attenuator on transmission side 2 regenerator 3 sound collector 4 echo canceller 5 received signal frequency band divider 6 collected signal frequency band divider 71~ 7N Subband echo processing unit 8 Transmission signal frequency band synthesis unit 11 Spectrum envelope estimation unit 12 Echo suppression filter calculation unit 13 Echo suppression filter 14 Decoding unit 15 Encoding unit A1 Reception signal A2 Transmission signal AA1 Reception code AA2 Transmission code 41 Pseudo-echo Generator 42 Subtractor
フロントページの続き Ｆターム(参考） 5D020 CC03 CC04 5K027 BB03 DD10 HH03 5K046 AA01 BA00 BB01 HH02 HH24 HH79 Continued on the front page F term (reference) 5D020 CC03 CC04 5K027 BB03 DD10 HH03 5K046 AA01 BA00 BB01 HH02 HH24 HH79
Claims (10)
器からの信号を送信側減衰部にて減衰して送信信号とす
る反響抑圧方法において、上記送信側減衰部は、 線形予測分析により受話信号から受話音声のスペクトル
包絡を求めるステップと、 このスペクトル包絡を平坦化する特性の反響抑圧フィル
タのフィルタ係数を算出するステップと、収音された音
響信号を上記反響抑圧フィルタに通過させて送信信号と
するステップと、を有することを特徴とする反響抑圧方
法。In a reverberation suppression method, a received signal is reproduced by a regenerator as an acoustic signal, and a signal from the sound pickup is attenuated by a transmission-side attenuator to become a transmission signal. Obtaining a spectrum envelope of the received voice from the received signal by analysis; calculating a filter coefficient of an echo suppression filter having a characteristic of flattening the spectrum envelope; and passing the collected sound signal through the echo suppression filter. And setting the signal as a transmission signal.
生器に出力される信号を参照して擬似反響信号を生成
し、収音器からの信号から上記擬似反響信号を差し引
き、その残差信号を送信側減衰部にて減衰して送信信号
とする反響抑圧方法において、上記送信側減衰部は、 線形予測分析により受話信号から受話音声のスペクトル
包絡を求めるステップと、このスペクトル包絡を平坦化
する特性の反響抑圧フィルタのフィルタ係数を算出する
ステップと、収音された音響信号を上記反響抑圧フィル
タに通過させて送信信号とするステップと、を有するこ
とを特徴とする反響抑圧方法。2. A reproduction apparatus for reproducing a reception signal into an acoustic signal, generating a pseudo echo signal by referring to a signal output to the reproduction apparatus, and subtracting the pseudo echo signal from a signal from the sound pickup device. In the echo suppression method in which the residual signal is attenuated by a transmission-side attenuator to become a transmission signal, the transmission-side attenuator obtains a spectrum envelope of a received voice from a received signal by linear prediction analysis, and An echo suppression method, comprising: calculating a filter coefficient of an echo suppression filter having characteristics to be flattened; and passing a collected sound signal through the echo suppression filter to form a transmission signal.
かにおいて、受話信号が高能率で符号化されて伝送され
る場合は、音声符号化のスペクトル包絡パラメータを記
録した符号と音源パラメータを記録した符号とを用い、
符号化入力からスペクトル包絡と音声信号を再生するこ
とを特徴とする反響抑圧方法。3. A method according to claim 1, wherein when the received signal is encoded and transmitted at a high efficiency, a code recording a spectrum envelope parameter of voice coding and a sound source are provided. Using the code that recorded the parameters,
A reverberation suppression method comprising reproducing a spectrum envelope and a speech signal from an encoded input.
かにおいて、線形予測分析により受話音声のスペクトル
包絡が全極型フィルタの係数として求められ、その分母
部分を取り出して得られるＦＩＲフィルタのゲインを調
整したフィルタを、スペクトル包絡を平坦化する反響抑
圧フィルタとして用いることを特徴とする反響抑圧方
法。4. A method according to claim 1, wherein the spectral envelope of the received voice is obtained as a coefficient of an all-pole filter by linear predictive analysis, and an FIR obtained by extracting a denominator portion thereof. An echo suppression method characterized by using a filter whose gain has been adjusted as an echo suppression filter for flattening a spectral envelope.
受話信号に分割する受話信号周波数帯域分割部と、 収音信号を複数の周波数帯域のサブバンド収音信号に分
割する収音信号周波数帯域分割部と、 各周波数帯域でサブバンド受話信号とサブバンド収音信
号から反響成分を減衰させてサブバンド送信信号を求め
る複数のサブバンド反響処理部と、 上記サブバンド反響処理部から出力されるサブバンド送
信信号を合成して送信信号とする送信信号周波数帯域合
成部と、 からなる反響抑圧方法において、 請求項１に記載の反響抑圧法もしくは請求項２記載の反
響抑圧法が上記複数のサブバンド反響処理部の何れかに
含まれることを特徴とする反響抑圧方法。5. A receiving signal frequency band dividing section for dividing a receiving signal into subband receiving signals of a plurality of frequency bands, and a collecting signal frequency band for dividing a collecting signal into subband collecting signals of a plurality of frequency bands. A dividing unit; a plurality of sub-band reverberation processing units for attenuating reverberation components from a sub-band reception signal and a sub-band pickup signal in each frequency band to obtain a sub-band transmission signal; A transmission signal frequency band synthesizing section for synthesizing a sub-band transmission signal to generate a transmission signal, wherein the echo suppression method according to claim 1 or the echo suppression method according to claim 2 comprises the plurality of sub-bands. A reverberation suppression method characterized by being included in any of the band reverberation processing units.
音器からの信号を送信側減衰部にて減衰して送信信号と
する反響抑圧装置において、 上記送信側減衰部は、 線形予測分析により受話信号から受話音声のスペクトル
包絡を求めるスペクトル包絡推定部と、 このスペクトル包絡推定部で求めたスペクトル包絡を平
坦化する特性の反響抑圧フィルタのフィルタ係数を算出
する反響抑圧フィルタ算出部と、 この反響抑圧フィルタ算出部で算出したフィルタ係数が
設定され、収音された音響信号を、この音響信号に含ま
れる反響成分を抑圧して通過させ送信信号とする反響抑
圧フィルタと、 によって構成されることを特徴とする反響抑圧装置。6. A reverberation suppressor for reproducing a reception signal into an acoustic signal by a regenerator and attenuating a signal from the sound pickup device by a transmission-side attenuator to form a transmission signal. A spectrum envelope estimator for obtaining a spectrum envelope of the received voice from the received signal by predictive analysis; and an echo suppression filter calculator for calculating a filter coefficient of an echo suppression filter having a characteristic of flattening the spectrum envelope obtained by the spectrum envelope estimator. A filter coefficient calculated by the echo suppression filter calculation unit is set, and the collected acoustic signal is a reverberation suppression filter that suppresses the reverberation component included in the acoustic signal and passes the signal to form a transmission signal. An echo suppression device characterized by the following.
生器に出力される信号を参照して擬似反響信号を生成
し、収音器からの信号から上記擬似反響信号を差し引
き、その残差信号を送信側減衰部にて減衰して送信信号
とする反響抑圧装置において、 上記送信側減衰部は、 線形予測分析により受話信号から受話音声のスペクトル
包絡を求めるスペクトル包絡推定部と、 このスペクトル包絡推定部で求めたスペクトル包絡を平
坦化する特性の反響抑圧フィルタのフィルタ係数を算出
する反響フィルタ算出部と、 この反響フィルタ算出部で算出したフィルタ係数が設定
され、収音された音響信号を、この音響信号に含まれる
反響成分を抑圧して通過させ送信信号とする反響抑圧フ
ィルタと、 によって構成されることを特徴とする反響抑圧装置。7. A received signal is reproduced by a reproducer as an acoustic signal, a pseudo echo signal is generated with reference to a signal output to the reproducer, and the pseudo echo signal is subtracted from a signal from the sound pickup device. In a reverberation suppression apparatus, the residual signal is attenuated by a transmission-side attenuator to be a transmission signal, wherein the transmission-side attenuator includes a spectrum envelope estimator for obtaining a spectrum envelope of a received voice from a received signal by linear prediction analysis; A reverberation filter calculator for calculating a filter coefficient of an echo suppression filter having a characteristic of flattening the spectrum envelope obtained by the spectrum envelope estimator; and a filter coefficient calculated by the reverberation filter calculator, and a collected sound signal. And a reverberation suppression filter for suppressing a reverberation component contained in the acoustic signal and passing the reverberation component to obtain a transmission signal.
かにおいて、受話信号が高能率で符号化されて伝送され
る場合は、音声符号化のスペクトル包絡パラメータを記
録した符号と音源パラメータを記録した符号とを用い、
符号化入力からスペクトル包絡と音声信号を再生する手
段を有する構成としたことを特徴とする反響抑圧装置。8. In a reverberation suppressing apparatus according to claim 6, when a received signal is encoded and transmitted with high efficiency, a code recording a spectrum envelope parameter of voice coding and a sound source are provided. Using the code that recorded the parameters,
A reverberation suppressing apparatus comprising means for reproducing a spectrum envelope and an audio signal from an encoded input.
かにおいて、線形予測分析により受話音声のスペクトル
包絡が全極型フィルタの係数として求められ、その分母
部分を取り出して得られるＦＩＲフィルタの利得を調整
したフィルタをスペクトル包絡を平坦化する反響抑圧フ
ィルタとして用いる構成としたことを特徴とする反響抑
圧装置。9. An echo canceller according to claim 6, wherein the spectral envelope of the received voice is obtained as a coefficient of an all-pole filter by linear prediction analysis, and an FIR obtained by extracting a denominator portion thereof. A reverberation suppression apparatus, wherein a filter having a gain adjusted for the filter is used as an echo suppression filter for flattening a spectral envelope.
ド受話信号に分割する受話信号周波数帯域分割部と、 収音信号を複数の周波数帯域のサブバンド収音信号に分
割する収音信号周波数帯域分割部と、 各周波数帯域でサブバンド受話信号とサブバンド収音信
号から反響成分を減衰させてサブバンド送信信号を求め
る複数のサブバンド反響処理部と、 上記サブバンド反響処理部から出力されるサブバンド送
信信号を合成して送信信号とする送信信号周波数帯域合
成部と、 からなる反響抑圧装置において、 請求項６に記載の反響抑圧装置もしくは請求項７に記載
の反響抑圧装置が上記複数のサブバンド反響処理部の何
れかに含まれることを特徴とする反響抑圧装置。10. A receiving signal frequency band dividing unit for dividing a receiving signal into sub-band receiving signals of a plurality of frequency bands, and a collecting signal frequency band for dividing a collecting signal into sub-band collecting signals of a plurality of frequency bands. A dividing unit; a plurality of sub-band reverberation processing units for attenuating reverberation components from a sub-band reception signal and a sub-band pickup signal in each frequency band to obtain a sub-band transmission signal; A transmission signal frequency band synthesizing unit for synthesizing a sub-band transmission signal to generate a transmission signal, wherein the echo suppression device according to claim 6 or the echo suppression device according to claim 7 comprises the plurality of echo suppression devices. A reverberation suppression device, which is included in any of the sub-band reverberation processing units.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2000246192A JP3724705B2 (en) | 2000-08-15 | 2000-08-15 | Echo suppression method / Echo suppression device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2000246192A JP3724705B2 (en) | 2000-08-15 | 2000-08-15 | Echo suppression method / Echo suppression device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JP2002064617A true JP2002064617A (en) | 2002-02-28 |
| JP3724705B2 JP3724705B2 (en) | 2005-12-07 |
Family
ID=18736559
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP2000246192A Expired - Fee Related JP3724705B2 (en) | 2000-08-15 | 2000-08-15 | Echo suppression method / Echo suppression device |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JP3724705B2 (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE10342307A1 (en) * | 2003-09-12 | 2005-04-28 | Siemens Ag | Method for interference suppression of transmitted audio signals between telecommunication terminals and detection method therefor |
| JP2006333062A (en) * | 2005-05-26 | 2006-12-07 | Nippon Telegr & Teleph Corp <Ntt> | Howling prevention method, device therefor, program therefor, and recording medium therefor |
| JP2011172165A (en) * | 2010-02-22 | 2011-09-01 | Oki Electric Industry Co Ltd | Echo cancellation apparatus, method and program |
| JP2012085314A (en) * | 2005-08-10 | 2012-04-26 | Clarity Technologies Inc | Method and system for clear signal capture |
| JP2015195510A (en) * | 2014-03-31 | 2015-11-05 | 沖電気工業株式会社 | Echo suppression device, echo suppression program and echo suppression method |
| JP2016503262A (en) * | 2012-12-21 | 2016-02-01 | マイクロソフト テクノロジー ライセンシング，エルエルシー | Echo suppression |
| WO2022195955A1 (en) * | 2021-03-16 | 2022-09-22 | パナソニック インテレクチュアル プロパティ コーポレーション オブ アメリカ | Echo suppressing device, echo suppressing method, and echo suppressing program |
-
2000
- 2000-08-15 JP JP2000246192A patent/JP3724705B2/en not_active Expired - Fee Related
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE10342307A1 (en) * | 2003-09-12 | 2005-04-28 | Siemens Ag | Method for interference suppression of transmitted audio signals between telecommunication terminals and detection method therefor |
| DE10342307B4 (en) * | 2003-09-12 | 2005-08-04 | Siemens Ag | Method for interference suppression of transmitted audio signals between telecommunication terminals |
| JP2006333062A (en) * | 2005-05-26 | 2006-12-07 | Nippon Telegr & Teleph Corp <Ntt> | Howling prevention method, device therefor, program therefor, and recording medium therefor |
| JP4551275B2 (en) * | 2005-05-26 | 2010-09-22 | 日本電信電話株式会社 | Howling prevention method, apparatus thereof, program thereof and recording medium thereof |
| JP2012085314A (en) * | 2005-08-10 | 2012-04-26 | Clarity Technologies Inc | Method and system for clear signal capture |
| JP2011172165A (en) * | 2010-02-22 | 2011-09-01 | Oki Electric Industry Co Ltd | Echo cancellation apparatus, method and program |
| JP2016503262A (en) * | 2012-12-21 | 2016-02-01 | マイクロソフト テクノロジー ライセンシング，エルエルシー | Echo suppression |
| JP2015195510A (en) * | 2014-03-31 | 2015-11-05 | 沖電気工業株式会社 | Echo suppression device, echo suppression program and echo suppression method |
| WO2022195955A1 (en) * | 2021-03-16 | 2022-09-22 | パナソニック インテレクチュアル プロパティ コーポレーション オブ アメリカ | Echo suppressing device, echo suppressing method, and echo suppressing program |
Also Published As
| Publication number | Publication date |
|---|---|
| JP3724705B2 (en) | 2005-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP4567655B2 (en) | Method and apparatus for suppressing background noise in audio signals, and corresponding apparatus with echo cancellation | |
| JP5671147B2 (en) | Echo suppression including modeling of late reverberation components | |
| KR100750440B1 (en) | Reverberation estimation and suppression system | |
| AU756511B2 (en) | Signal noise reduction by spectral subtraction using linear convolution and causal filtering | |
| US9992572B2 (en) | Dereverberation system for use in a signal processing apparatus | |
| US7620172B2 (en) | Method and system for eliminating noises and echo in voice signals | |
| EP1080463B1 (en) | Signal noise reduction by spectral subtraction using spectrum dependent exponential gain function averaging | |
| US20100246844A1 (en) | Method for Determining a Signal Component for Reducing Noise in an Input Signal | |
| US20090238373A1 (en) | System and method for envelope-based acoustic echo cancellation | |
| US8306821B2 (en) | Sub-band periodic signal enhancement system | |
| KR20040030817A (en) | Echo canceller having nonlinear echo suppressor for harmonics calculations | |
| CN110956975B (en) | Echo cancellation method and device | |
| JP5130895B2 (en) | Audio processing apparatus, audio processing system, audio processing program, and audio processing method | |
| JP3507020B2 (en) | Echo suppression method, echo suppression device, and echo suppression program storage medium | |
| JP4105681B2 (en) | Echo suppress method, echo suppressor, echo suppressor program, loss control method on communication path, loss control device on communication path, loss control program on communication path, recording medium | |
| JP2003284183A (en) | Echo suppression apparatus, echo suppression method, and program | |
| JP3724705B2 (en) | Echo suppression method / Echo suppression device | |
| JP2003188776A (en) | Acoustic echo erasing method and device, and acoustic echo erasure program | |
| JPH11265199A (en) | Voice transmitter | |
| WO2020023856A1 (en) | Forced gap insertion for pervasive listening | |
| CN115579016B (en) | Method and system for eliminating acoustic echo | |
| JP2003309493A (en) | Method, device and program for reducing echo | |
| JPH08223677A (en) | Telephone transmitter | |
| JP4253232B2 (en) | Noise suppression method, noise suppression device, noise suppression program | |
| JP4594854B2 (en) | Voice switch method, voice switch device, voice switch program, and recording medium recording the program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| A977 | Report on retrieval |
Effective date: 20040524 Free format text: JAPANESE INTERMEDIATE CODE: A971007 |
|
| A131 | Notification of reasons for refusal |
Free format text: JAPANESE INTERMEDIATE CODE: A131 Effective date: 20040615 |
|
| A521 | Written amendment |
Effective date: 20040730 Free format text: JAPANESE INTERMEDIATE CODE: A523 |
|
| 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: 20050830 |
|
| RD03 | Notification of appointment of power of attorney |
Effective date: 20050915 Free format text: JAPANESE INTERMEDIATE CODE: A7423 |
|
| A61 | First payment of annual fees (during grant procedure) |
Effective date: 20050915 Free format text: JAPANESE INTERMEDIATE CODE: A61 |
|
| R150 | Certificate of patent (=grant) or registration of utility model |
Free format text: JAPANESE INTERMEDIATE CODE: R150 |
|
| FPAY | Renewal fee payment (prs date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20080930 Year of fee payment: 3 |
|
| FPAY | Renewal fee payment (prs date is renewal date of database) |
Year of fee payment: 4 Free format text: PAYMENT UNTIL: 20090930 |
|
| FPAY | Renewal fee payment (prs date is renewal date of database) |
Year of fee payment: 4 Free format text: PAYMENT UNTIL: 20090930 |
|
| FPAY | Renewal fee payment (prs date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20100930 Year of fee payment: 5 |
|
| FPAY | Renewal fee payment (prs date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20100930 Year of fee payment: 5 |
|
| FPAY | Renewal fee payment (prs date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20110930 Year of fee payment: 6 |
|
| FPAY | Renewal fee payment (prs date is renewal date of database) |
Year of fee payment: 7 Free format text: PAYMENT UNTIL: 20120930 |
|
| FPAY | Renewal fee payment (prs date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20130930 Year of fee payment: 8 |
|
| LAPS | Cancellation because of no payment of annual fees |