JP3621718B2 - Vehicle interior noise reduction device - Google Patents

Description

となる。
【００２７】
尚、図１中、符号Ｃはエンジン１の振動騒音に対する車体伝達特性を示す。
【００２８】
次に、上記構成による実施例の作用について説明する。
まず、エンジンの振動騒音は、エンジン１から図示しないマウント等を伝達して車内音となり、また、吸気や排気の音等も車室内に伝播する。これらのエンジン関連振動騒音は、周波数領域では、いずれも０．５×ｎ（ｎ：整数）次成分の周波数スペクトルにより主に構成されており、各々の振動源に対する車体伝達特性Ｃが乗ぜられて受聴点（例えばドライバーの耳に近接する位置）に達する。
【００２９】
一方、エンジン１のイグニッションコイル（図示せず）へのイグニッションパルス信号（Ｉｇ パルス信号）は、入力信号変換回路２に入力され、波形成形回路２ａと間引回路２ｂにより、エンジン回転に同期してエンジン２回転で１パルスで、エンジン回転の０．５×ｎ（ｎ：整数）次成分の周波数からなる信号に成形・間引されて、騒音振動源信号（プライマリソースＰｓ ）として、適応フィルタ３、スピーカ／マイク間伝達特性補償回路（以下「ＣＭＮ０ 回路」と略称）４およびノイズ防止回路５に出力される。
【００３０】
上記適応フィルタ３に入力されたプライマリソースＰｓ は、この適応フィルタ３のフィルタ係数Ｗ（ｎ） との畳み込み積和により、振動騒音を相殺するキャンセル信号として、Ｄ／Ａ変換器７に出力され、図示しないフィルタ回路およびアンプ回路（ＡＭＰ回路）８を介して、スピーカ９に出力され、このスピーカ９から上記受聴点における振動騒音に対する相殺音として出力される。このとき、上記相殺音は、スピーカ／マイク間伝達特性ＣＭＮを受けて上記受聴点に達する。
【００３１】
このため、上記受聴点では、上記エンジン関連の振動騒音と上記相殺音とが干渉して振動騒音が低減させられると同時に、上記受聴点の近傍に配設されているマイク１０により、振動騒音と相殺音との干渉の結果が検出され、エラー信号ｅ（ｉ）ｎとして、アンプ回路（ＡＭＰ回路）１１、フィルタ回路（図示せず）およびＡ／Ｄ変換器１２を介して上記ノイズ防止回路５に入力される。
【００３２】
上記ノイズ防止回路５では、上記プライマリソースＰｓ がトリガ信号生成部５ａに入力され、このプライマリソースＰｓ のパルスをトリガパルスとして生成して、エラー信号記憶部５ｂとエラー信号比較・判定部５ｃとに出力する。
【００３３】
そして、上記エラー信号ｅ（ｉ）ｎがエラー信号記憶部５ｂに１周期分にわたって記憶されるとともに、エラー信号比較・判定部５ｃに入力されて、上記エラー信号記憶部５ｂに記憶されている前回のエラー信号ｅ（ｉ）ｎ−１と上記マイク１０から入力されたエラー信号ｅ（ｉ）ｎとの差（｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜）が求められ、予め実験等により求め設定しておいた値αと比較される。
【００３４】
次いで、エラー信号設定部５ｄで、上記エラー信号比較・判定部５ｃでの比較結果に基づき、｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜≦αの場合は、上記マイク１０から入力されたエラー信号ｅ（ｉ）ｎには周期音以外のノイズ成分（ランダム信号；人の話し声や音楽あるいは突発的な音等）が含まれないと判断して上記ＬＭＳ演算回路６に出力し、また、｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜＞αの場合は、上記前回のエラー信号ｅ（ｉ）ｎ−１、あるいは、上記マイク１０から入力されたエラー信号ｅ（ｉ）ｎのどちらかにノイズ成分が含まれると判断して、上記ｅ（ｉ）ｎ−１と上記ｅ（ｉ）ｎの他に、上記ｅ（ｉ）ｎ−１と上記ｅ（ｉ）ｎの平均値（（ｅ（ｉ）ｎ−１＋ｅ（ｉ）ｎ）／２）を求め、これら３つの信号（ｅ（ｉ）ｎ−１，ｅ（ｉ）ｎ，（ｅ（ｉ）ｎ−１＋ｅ（ｉ）ｎ）／２）の中からエラー量（振幅の絶対値）の最も小さい信号を選択して上記ＬＭＳ演算回路６に出力する。
【００３５】
そして、上記ＬＭＳ演算回路６で、上記ノイズ防止回路５からの信号と、上記ＣＭＮ０ 回路４で補償されたプライマリソースＰｓ とから、周知のＬＭＳアルゴリズムにより前記適応フィルタ３のフィルタ係数Ｗ（ｎ） の修正量を求め、フィルタ係数Ｗ（ｎ） を更新する。
【００３６】
このため、本第一実施例によれば、例えば、前回のエラー信号が略ゼロで突発入力があった場合、いかなる突発入力も完全に阻止することができる。同様に、突発入力が前回のエラー信号に比較して充分大きい場合にも完全に阻止することができる。
【００３７】
また、周期毎に交番する大きな入力があっても、前回のエラー信号と今回のエラー信号との平均値が適応されるようになるため、大きな入力が入ることはない。
【００３８】
さらに、いかに大きなエラー信号であっても、２周期以上（２データ以上）連続して同様な信号となった場合には、ＬＭＳ演算回路に入力されることになり、追従性の劣化を最小限に抑えることができる。
【００３９】
また、適応の遅い成分が充分消音された（適応された）のち、騒音が急に除去された場合に残る相殺音の減衰に対しては、騒音除去直後のエラー信号は阻止されてしまうが、その次のデータからは前回のデータとほぼ同様のエラー信号（或いは、より小さいエラー信号）として検出されるので、今回のエラー信号がそのままエラー信号として用いられることになり、減衰速度の劣化は最小限に抑えることができる。
【００４０】
尚、本第一実施例では、プライマリソースＰｓ としてＩｇ パルスを用いるように構成しているが、他のエンジン関連の振動騒音と相関の高い信号（例えば、燃料噴射パルスＴｉ 等）をプライマリソースＰｓ としても良い。
【００４１】
また、本第一実施例では、１チャンネル（マイク１個、スピーカ１個）のＬＭＳアルゴリズムを利用した騒音低減装置の例について説明したが、ＬＭＳアルゴリズムを多チャンネルに拡大したＭＥＦＸ−ＬＭＳ（Ｍｕｌｔｉｐｌｅ Ｅｒｒｏｒ Ｆｉｌｔｅｒｅｄ Ｘ−ＬＭＳ）アルゴリズムを利用した車室内騒音低減装置（例えば、マイク４個、スピーカ４個等の装置）についても適用可能である。
【００４２】
次いで、図４および図５は本発明の第二実施例を示し、図４は車室内騒音低減装置のシステム概略図、図５はノイズ防止回路の説明図である。尚、この第二実施例は、ノイズ防止回路のエラー信号比較・判定部での比較結果に基づき、適応フィルタのフィルタ係数更新式のステップサイズを可変させるようにしたものであり、前記第一実施例と同じ部分には同一符号を記しその説明は省略する。
【００４３】
図４において、符号２１は車室内騒音低減装置を示し、マイク１０にて検出された騒音低減状態を示す誤差信号（エラー信号）は、アンプ回路（ＡＭＰ回路）１１、フィルタ回路（図示せず）およびＡ／Ｄ変換器１２を介して、ＬＭＳ演算回路６とノイズ防止回路２５とに入力されるようになっている。
【００４４】
また、入力信号変換回路２からのプライマリソースＰｓ も、上記ノイズ防止回路２５に入力される。
【００４５】
上記ノイズ防止回路２５は、トリガ信号生成部５ａと、エラー信号記憶部５ｂと、エラー信号比較・判定部５ｃと、ステップサイズ設定部２５ｄとから主に構成されている。
【００４６】
上記トリガ信号生成部５ａ，エラー信号記憶部５ｂ，エラー信号比較・判定部５ｃは、前記第一実施例と同様に、上記トリガ信号生成部５ａには、上記プライマリソースＰｓ が入力され、このプライマリソースＰｓ のパルスを、上記エラー信号記憶部５ｂと上記エラー信号比較・判定部５ｃで行う信号処理のトリガパルスとして生成する回路部で、上記エラー信号記憶部５ｂは、ＲＡＭ回路で、上記マイク１０から入力されたエラー信号ｅ（ｉ）ｎを、１周期分にわたって（１周期分を１データとして）記憶する回路部に形成されている。また、上記エラー信号比較・判定部５ｃは、上記エラー信号記憶部５ｂに記憶されている前回のエラー信号（前回のデータ）ｅ（ｉ）ｎ−１と上記マイク１０から入力されたエラー信号（今回のデータ）ｅ（ｉ）ｎとの差（｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜）を求め、予め実験等により求め設定しておいた値αと比較する回路部である。
【００４７】
さらに、上記ステップサイズ設定部２５ｄは、上記エラー信号比較・判定部５ｃでの比較結果に基づき、｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜≦αの場合は、上記マイク１０から入力されたエラー信号ｅ（ｉ）ｎには周期音以外のノイズ成分（ランダム信号；人の話し声や音楽あるいは突発的な音等）が含まれないと判断して上記ＬＭＳ演算回路６に対してステップサイズμの変更は行わず、また、｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜＞αの場合は、上記今回のデータｅ（ｉ）ｎの中の｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜＞αとなるサンプル数をカウントして、この割合Ｓｐ を求め、この割合が予め実験等で設定しておいた割合β以上の際に、上記ＬＭＳ演算回路６に対して適応フィルタ３のフィルタ係数Ｗ（ｎ） の更新式中のステップサイズμを小さく設定する回路部となっている。ここで、Ｓｐ ≧βの際は、上記割合をＳｐ ，小さく設定したステップサイズをμ’ とすると、
Ｓｐ ＝｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜＞αのサンプル数／ｅ（ｉ）ｎの全サンプル数…（２）
また、ステップサイズをμ’ は、１より大きい値γを用いて、
μ’ ＝μ×（１／γ） …（３）
で与えるものとし、（但し、｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜＞αのサンプルにのみ適用）｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜＞αのサンプルでの上記適応フィルタ３のフィルタ係数Ｗ（ｎ） 更新は、前記（１）式から、

の更新式で行われる。
【００４８】
一方、上記ステップサイズ設定部２５ｄで、上記エラー信号比較・判定部５ｃでの比較結果が、｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜＞αの場合で、Ｓｐ ＜βの場合は、ステップサイズμの変更は行わない。
【００４９】
尚、ステップサイズμ’ は、他の式によって求めても良く、例えば、定数Ｋ，μ０ を用いて、
μ’ ＝Ｋ（１−Ｓｐ ）＋μ０
或いは、コンピュータ処理を考慮して、
μ’ ＝μ／（２）^{（Ｓ ｐ ／β）} …（３）’
のように、Ｓｐ により連続的に与えられるようにしても良い。
【００５０】
また、Ｓｐ は実施例では（２）式のように｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜＞αのサンプル数の比率としたが、｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜の値をｅ（ｉ）ｎの全サンプル分足し合わせた値、即ち、
Ｓｐ ＝Σ｜ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ｜としても良いし、
Ｓｐ ＝Σ（ｅ（ｉ）ｎ−１−ｅ（ｉ）ｎ）^２ としても良い。
【００５１】
すなわち、本第二実施例では、エラー信号へのノイズ成分の混入を、このノイズ成分混入の割合で判定し、ノイズ成分混入の割合が大きいと判定しても、適応速度は全データに対して遅くするのではなく、ノイズ成分混入のサンプルに対してのみ遅くするようしている。このため、ノイズ成分のない消音対象とする周期的な成分に対する適応は劣化することがなく、消音性能を高く保ちながら、ノイズ成分の除去が行えるのである。
【００５２】
また、エラー信号へのノイズ成分の混入を、ノイズ成分混入の割合で判定するようにしているため、ノイズ成分の混入量に応じた適応速度のコントロールが行われるので長い時間連続したノイズ成分に対しても、確実に入力を阻止することが可能である。
【００６２】
【発明の効果】
以上、説明したように本発明によれば、例え、人の話し声や音楽あるいは突発的な音等のランダム信号が発生しても、このランダム信号による残響を、消音性能が劣化することなく抑えることが可能となる。
【図面の簡単な説明】
【図１】本発明の第一実施例による車室内騒音低減装置のシステム概略図
【図２】本発明の第一実施例による入力信号変換回路の説明図
【図３】本発明の第一実施例によるノイズ防止回路の説明図
【図４】本発明の第二実施例による車室内騒音低減装置のシステム概略図
【図５】本発明の第二実施例による入力信号変換回路の説明図
【符号の説明】
１ エンジン
３ 適応フィルタ（キャンセル信号合成手段）
５ ノイズ防止回路（ノイズ防止手段）
６ ＬＭＳ演算回路（フィルタ係数更新手段）
９ スピーカ（相殺音発生手段）
１０ マイク（誤差信号検出手段）
２５ ノイズ防止回路（ノイズ防止手段）
ｅ(i)n エラー信号（誤差信号）
ｅ(i)n-1 前回のデータ
α 設定値
β 設定割合
Ｐs プライマリソース（騒音振動源信号）
Ｓp ｜ｅ(i)n-1−ｅ(i)n｜＞αとなるサンプル数の割合
Ｗ(n) フィルタ係数[0001]
[Industrial application fields]
The present invention relates to a vehicle interior noise reduction device that reduces vehicle interior noise generated mainly due to engine vibration noise by interfering with canceling sound.
[0002]
[Prior art]
Various technologies have been proposed to reduce vehicle interior noise by generating sound (cancelling sound) from the sound source that has the same amplitude and opposite phase as the noise caused by engine vibration noise as the main factor. Yes.
[0003]
Recently, as shown in, for example, Japanese Patent Laid-Open No. 3-178845, an LMS (Least Mean Square) algorithm (a calculation formula for obtaining a filter coefficient of an adaptive filter has been simplified. Therefore, there is a vehicle interior noise reduction device using a multi-filter filtered X-LMS (MEFX-LMS) algorithm that is obtained by expanding the LMS algorithm to multiple channels. It has been proposed and partly put into practical use. In the vehicle interior noise reduction device using this LMS algorithm, when the vehicle interior noise generated mainly due to engine vibration is silenced, a signal highly correlated with engine vibration is detected as a noise vibration source signal (primary source). The canceling sound for the noise is synthesized from the primary source by the adaptive filter and generated from the speaker. Then, the noise reduction state at the listening point is detected by the microphone as an error signal, and the filter coefficient of the adaptive filter is updated from the error signal and the primary source by the LMS algorithm so that the noise reduction at the listening point becomes an optimum value. It has become.
[0004]
[Problems to be solved by the invention]
By the way, in the vehicle interior noise reduction device for periodic sounds using the above-described LMS algorithm or MEFX-LMS algorithm, the noise reduction performance is generally improved by increasing the adaptive speed of the filter coefficient update of the adaptive filter. Since it is also applied to noise components (random signals; human speech, music or sudden sounds) other than the periodic sound to be muffled, a periodic canceling sound for this noise component can be emitted from the speaker. May feel reverberant and give passengers a sense of incongruity.
[0005]
Further, if the adaptive speed is set to be slow, the reverberation as described above can be reduced, but there is a problem that the silencing performance is deteriorated.
[0006]
The present invention has been made in view of the above circumstances. For example, even if a random signal such as a human voice or music or a sudden sound is generated, the reverberation due to the random signal is not deteriorated. An object of the present invention is to provide a vehicle interior noise reduction device that can be suppressed.
[0007]
[Means for Solving the Problems]
In order to achieve the above object, a first vehicle interior noise reduction apparatus according to the present invention comprises a cancel signal combining means for combining a noise vibration source signal having a high correlation with engine vibration as a cancel signal by an adaptive filter, and the cancel signal for noise. A canceling sound generating means for generating a canceling sound, an error signal detecting means for detecting a noise reduction state at the listening point as an error signal, at least one period or more of the error signal is stored as one data, and the previous data and the above When a difference from the error signal is obtained and the difference is equal to or smaller than a preset value, the error signal is output, whereas when the difference between the previous data and the error signal is larger than the set value, the previous data, the error signal and an average processed signal with the weighting of the error signal and the preset value and the previous data Noise prevention means for selecting and outputting the signal having the smallest absolute value from the above, and filter coefficient updating means for updating the filter coefficient of the adaptive filter based on the noise vibration source signal and the signal from the noise prevention means. It is a thing.
[0008]
In order to achieve the above object, a second vehicle interior noise reduction apparatus according to the present invention comprises a cancel signal combining means for combining a noise vibration source signal having a high correlation with engine vibration as a cancel signal by an adaptive filter, and the cancel signal. A canceling sound generating means for generating a canceling sound for noise, an error signal detecting means for detecting a noise reduction state at the listening point as an error signal, and storing at least one period of the error signal as one data, and the previous data If the difference is larger than a preset value , the difference between the previous data and the current error signal in the current error signal is larger than the set value. obtains the number of samples made, determine the percentage of the total number of samples of the current error signal of this number of samples, dividing this ratio preset A noise preventing means for setting a slow update rate of the filter coefficient of the adaptive filter for a signal portion in which the difference between the previous data and the error signal is greater than the set value, and the noise vibration source signal And a filter coefficient updating means for updating the filter coefficient of the adaptive filter based on the error signal.
[0010]
[Operation]
In the first vehicle interior noise reduction device, first, when noise occurs in the vehicle interior due mainly to engine vibration noise, the noise signal source having a high correlation with the engine vibration is canceled by the adaptive filter when the noise is generated in the vehicle interior. The cancel signal is synthesized as a signal, and the cancel signal is generated as a cancel sound for noise from the cancel sound generating means.
Next, the error signal detection means detects the noise reduction state at the listening point as an error signal, and the noise prevention means stores at least one cycle of the error signal as one data, and the previous data and the error signal. If the difference is less than a preset value, the error signal is output. On the other hand, if the difference between the previous data and the error signal is greater than the set value, the previous data is output. When the above error signal, selects and outputs a smaller signal the least absolute value from the average processed signal with the weighting of the error signal and the preset value and the previous data.
Then, the filter coefficient updating means updates the filter coefficient of the adaptive filter based on the noise vibration source signal and the signal from the noise prevention means.
[0011]
In the second vehicle interior noise reduction device, first, when noise is generated in the vehicle interior due mainly to engine vibration noise, a noise signal source having a high correlation with engine vibration is applied to the adaptive filter by the cancel signal synthesis means. The cancel signal is synthesized as a cancel signal, and the cancel signal is generated as a cancel sound for the noise from the cancel sound generating means.
Next, the error signal detection means detects the noise reduction state at the listening point as an error signal, and the noise prevention means stores at least one cycle of the error signal as one data, and the previous data and the error signal. If this difference is larger than the preset value, the number of samples in which the difference between the previous data and the current error signal in the current error signal is greater than the set value is obtained. , The ratio of the number of samples to the total number of samples of the current error signal, and when this ratio is greater than or equal to a preset ratio, a signal in which the difference between the previous data and the error signal is greater than the set value The update speed of the filter coefficient of the adaptive filter for the portion is set to be slow.
Then, the filter coefficient updating means updates the filter coefficient of the adaptive filter based on the noise vibration source signal and the error signal.
[0013]
【Example】
Embodiments of the present invention will be described below with reference to the drawings.
1 to 3 show a first embodiment of the present invention, FIG. 1 is a system schematic diagram of a vehicle interior noise reduction device, FIG. 2 is an explanatory diagram of an input signal conversion circuit, and FIG. 3 is an explanatory diagram of a noise prevention circuit. is there.
[0014]
In FIG. 1, reference numeral 1 denotes a four-cycle engine. An ignition pulse signal (Ig pulse signal) to an ignition coil (not shown) of the engine 1 is also output to the input signal conversion circuit 2.
[0015]
As shown in FIG. 2, the input signal conversion circuit 2 includes a waveform shaping circuit 2a and a thinning circuit 2b. The Ig pulse signal input to the input signal conversion circuit 2 is used for engine rotation. Synchronously with one pulse of two engine revolutions, it is shaped and thinned into a signal consisting of the frequency of the 0.5 × n (n: integer) order component of the engine revolution, and as a noise vibration source signal (primary source Ps), The signal is output to an adaptive filter 3 as a cancel signal synthesis means, a speaker / microphone transfer characteristic compensation circuit (hereinafter abbreviated as “CMN0 circuit”) 4 and a noise prevention circuit 5 as noise prevention means.
[0016]
This is because the vibration noise related to the 4-cycle engine is the vibration noise with 2 rotations of the engine as one cycle in order to complete 4 strokes of suction, compression, explosion and exhaust in 2 rotations (720 ° C). In the frequency domain, the 0.5-order component of the engine rotation is the fundamental wave, and the spectrum is mainly composed of the higher-order component (consisting of 0.5 × n (n: integer) order component). Because). Therefore, by shaping and processing the Ig pulse signal as described above, it is possible to obtain the primary source Ps having a very high correlation with the vibration noise to be silenced.
[0017]
The adaptive filter 3 is an FIR (Finite Impulse Response) filter having a filter coefficient W (n) that can be updated by the LMS arithmetic circuit 6 as a filter coefficient updating unit, and has a predetermined number of taps (for example, 512 taps). Is formed. The primary source Ps input to the adaptive filter 3 is convolved and summed with the filter coefficient W (n), and is output to the D / A converter 7 as a cancel signal, and a filter circuit and an amplifier circuit (AMP) (not shown). A canceling sound is generated from a speaker 9 as a canceling sound generating means via a circuit 8.
[0018]
The speaker 9 is disposed, for example, on a front door or the like in a vehicle (not shown), and a microphone as an error signal detection means at a listening point in the vehicle (for example, a position close to the ear position of a passenger on the driver's seat). 10 is provided.
[0019]
An error signal indicating a noise reduction state detected by the microphone 10 (a signal indicating a result of interference between the canceling sound and engine-related vibration noise; an error signal) includes an amplifier circuit (AMP circuit) 11 and a filter circuit (FIG. And the noise prevention circuit 5 through the A / D converter 12.
[0020]
As shown in FIG. 3, the noise prevention circuit 5 mainly includes a trigger signal generation unit 5a, an error signal storage unit 5b, an error signal comparison / determination unit 5c, and an error signal setting unit 5d. .
[0021]
In other words, the primary source Ps is input to the trigger signal generation unit 5a, and the pulse of the primary source Ps is used as a trigger pulse for signal processing performed by the error signal storage unit 5b and the error signal comparison / determination unit 5c. It is a circuit part to generate.
[0022]
The error signal storage unit 5b is a RAM circuit, and is formed in a circuit unit that stores the error signal e (i) n input from the microphone 10 for one period (one period as one data). ing. The sampling frequency in the error signal storage unit 5b is preferably as high as possible (for example, 3 kHz). However, in order to save the amount of memory if the silencing performance of the silencing system can be ensured, for example, at 1 kHz Sampling (sampling every several samples) is also possible. Further, instead of simply sampling every several samples, it may be data consisting of an average value of several samples. In this embodiment, the error signal e (i) n is stored for one period. However, a value of one period or more may be stored as one data.
[0023]
Further, the error signal comparison / determination unit 5c includes the previous error signal (previous data) e (i) n-1 stored in the error signal storage unit 5b and the error signal input from the microphone 10 ( A circuit unit for obtaining a difference (| e (i) n-1-e (i) n |) from e (i) n and comparing with a value α obtained and set in advance by an experiment or the like; It has become.
[0024]
Further, the error signal setting unit 5d determines from the microphone 10 when | e (i) n-1-e (i) n | ≦ α based on the comparison result in the error signal comparison / determination unit 5c. It is determined that the input error signal e (i) n does not contain noise components other than periodic sounds (random signals; human speech, music, sudden sounds, etc.), and outputs the result to the LMS arithmetic circuit 6. It is a circuit part. When | e (i) n-1-e (i) n |> α, the previous error signal e (i) n-1 or the error signal e (i) input from the microphone 10 is used. ) It is determined that a noise component is included in one of n, and in addition to e (i) n-1 and e (i) n, e (i) n-1 and e (i) n The average value of (e (i) n-1 + e (i) n) / 2) is obtained, and these three signals (e (i) n-1, e (i) n, (e (i) n-1 + e) are obtained. (I) A circuit unit that selects a signal with the smallest error amount (absolute amplitude value) from n) / 2) and outputs the selected signal to the LMS arithmetic circuit 6. This selection is based on the above e (i) n-1, the above e (i) n, and other signals (for example, exponents) obtained by processing this e (i) n based on the above e (i) n-1. Average value: If the previous exponent average value is Pxn-1, and the current exponent average value is Pxn, then Pxn = ((N-1) Pxn-1 + e (i) n) / N; N is a constant of 1 or more) You may make it choose from.
[0025]
On the other hand, in the CMN0 circuit 4, a speaker / microphone transfer characteristic CMN is set in advance by approximation with a finite impulse response (as a compensation coefficient CMN0), and the compensation coefficient CMN0 is applied to the input primary source Ps. This circuit compensates by multiplication (convolution product sum) and outputs a signal to the LMS arithmetic circuit 6.
[0026]
In the LMS arithmetic circuit 6, the signal from the noise prevention circuit 5 and the primary source Ps compensated (convolved and summed) by the CMN0 circuit 4 are used for the adaptive filter 3 by a known LMS algorithm. The correction amount of the filter coefficient W (n) is obtained, and the filter coefficient W (n) is updated. Here, the updated i-th filter coefficient is Wi (n + 1), the i-th filter coefficient to be updated is Wi (n), the step size is μ, and from the noise prevention circuit 5 input to the LMS arithmetic circuit 6. Is E (n), the jth coefficient of the compensation coefficient CMN0 (J tap) is CMjN0, and the value of the primary source Ps is x (n).

It becomes.
[0027]
In FIG. 1, symbol C indicates a vehicle body transmission characteristic with respect to vibration noise of the engine 1.
[0028]
Next, the operation of the embodiment having the above configuration will be described.
First, the vibration noise of the engine is transmitted from the engine 1 through a mount (not shown) or the like to be generated in the vehicle interior, and the intake and exhaust sounds are also transmitted into the vehicle interior. These engine-related vibration noises are mainly composed of a frequency spectrum of 0.5 × n (n: integer) order component in the frequency domain, and multiplied by the vehicle body transfer characteristic C for each vibration source. A listening point (for example, a position close to the driver's ear) is reached.
[0029]
On the other hand, an ignition pulse signal (Ig pulse signal) to an ignition coil (not shown) of the engine 1 is input to the input signal conversion circuit 2 and synchronized with the engine rotation by the waveform shaping circuit 2a and the thinning circuit 2b. The adaptive filter 3 is formed as a noise vibration source signal (primary source Ps) by shaping and decimating a signal having a frequency of 0.5 × n (n: integer) order component of engine rotation in one pulse with two engine rotations. Are output to the speaker / microphone transfer characteristic compensation circuit (hereinafter abbreviated as “CMN0 circuit”) 4 and the noise prevention circuit 5.
[0030]
The primary source Ps input to the adaptive filter 3 is output to the D / A converter 7 as a cancel signal that cancels vibration noise by the convolution product sum with the filter coefficient W (n) of the adaptive filter 3. The sound is output to a speaker 9 via a filter circuit and an amplifier circuit (AMP circuit) 8 (not shown), and output from the speaker 9 as a canceling sound for vibration noise at the listening point. At this time, the canceling sound receives the speaker / microphone transfer characteristic CMN and reaches the listening point.
[0031]
For this reason, at the listening point, the vibration noise associated with the engine-related vibration noise and the canceling sound are reduced, and at the same time, the vibration noise is reduced by the microphone 10 disposed in the vicinity of the listening point. The result of interference with the canceling sound is detected, and the noise prevention circuit 5 is passed through the amplifier circuit (AMP circuit) 11, the filter circuit (not shown) and the A / D converter 12 as the error signal e (i) n. Is input.
[0032]
In the noise prevention circuit 5, the primary source Ps is input to the trigger signal generation unit 5a, and a pulse of the primary source Ps is generated as a trigger pulse to be sent to the error signal storage unit 5b and the error signal comparison / determination unit 5c. Output.
[0033]
The previous error signal e (i) n is stored in the error signal storage unit 5b for one period, and is input to the error signal comparison / determination unit 5c and stored in the error signal storage unit 5b. Difference (| e (i) n-1-e (i) n |) between the error signal e (i) n-1 and the error signal e (i) n input from the microphone 10 is obtained in advance. It is compared with the value α obtained and set by experiments.
[0034]
Next, the error signal setting unit 5d inputs from the microphone 10 when | e (i) n-1-e (i) n | ≦ α based on the comparison result in the error signal comparison / determination unit 5c. It is determined that the error signal e (i) n is free of noise components other than periodic sounds (random signals; human speech, music, sudden sounds, etc.) and is output to the LMS arithmetic circuit 6. When | e (i) n-1-e (i) n |> α, the previous error signal e (i) n-1 or the error signal e (i) input from the microphone 10 is used. ) It is determined that a noise component is included in one of n, and in addition to e (i) n-1 and e (i) n, e (i) n-1 and e (i) n The average value of (e (i) n-1 + e (i) n) / 2) is obtained, and these three signals (e (i) n-1, e (i) A signal having the smallest error amount (absolute value of amplitude) is selected from n, (e (i) n-1 + e (i) n) / 2) and output to the LMS arithmetic circuit 6.
[0035]
Then, the LMS arithmetic circuit 6 uses the signal from the noise prevention circuit 5 and the primary source Ps compensated by the CMN0 circuit 4 to calculate the filter coefficient W (n) of the adaptive filter 3 by a known LMS algorithm. The correction amount is obtained and the filter coefficient W (n) is updated.
[0036]
For this reason, according to the first embodiment, for example, when the previous error signal is substantially zero and there is a sudden input, any sudden input can be completely prevented. Similarly, even when the sudden input is sufficiently larger than the previous error signal, it can be completely prevented.
[0037]
Also, even if there is a large input that alternates every cycle, the average value of the previous error signal and the current error signal is adapted, so that there is no large input.
[0038]
Furthermore, no matter how large the error signal is, if it becomes the same signal continuously for 2 cycles or more (2 data or more), it will be input to the LMS arithmetic circuit, and the follow-up deterioration will be minimized. Can be suppressed.
[0039]
In addition, after the slow-adapted component is sufficiently silenced (adapted), the error signal immediately after the noise removal is prevented against the attenuation of the canceling sound that remains when the noise is suddenly removed. From the next data, it is detected as an error signal that is almost the same as the previous data (or a smaller error signal), so the current error signal is used as it is as an error signal, and degradation of the attenuation rate is minimal. To the limit.
[0040]
In the first embodiment, an Ig pulse is used as the primary source Ps. However, a signal (for example, fuel injection pulse Ti or the like) highly correlated with other engine-related vibration noise is used as the primary source Ps. It is also good.
[0041]
In the first embodiment, an example of the noise reduction device using the LMS algorithm of one channel (one microphone and one speaker) has been described. However, MEFX-LMS (Multiple Error) in which the LMS algorithm is expanded to multiple channels. The present invention is also applicable to a vehicle interior noise reduction device (for example, a device having four microphones, four speakers, etc.) using a Filtered X-LMS algorithm.
[0042]
4 and 5 show a second embodiment of the present invention, FIG. 4 is a system schematic diagram of a vehicle interior noise reduction device, and FIG. 5 is an explanatory diagram of a noise prevention circuit. In the second embodiment, the step size of the filter coefficient updating formula of the adaptive filter is made variable based on the comparison result in the error signal comparison / determination unit of the noise prevention circuit. The same parts as those in the example are denoted by the same reference numerals and the description thereof is omitted.
[0043]
In FIG. 4, reference numeral 21 denotes a vehicle interior noise reduction device, and an error signal (error signal) indicating a noise reduction state detected by the microphone 10 is an amplifier circuit (AMP circuit) 11 and a filter circuit (not shown). The signal is input to the LMS arithmetic circuit 6 and the noise prevention circuit 25 via the A / D converter 12.
[0044]
The primary source Ps from the input signal conversion circuit 2 is also input to the noise prevention circuit 25.
[0045]
The noise prevention circuit 25 mainly includes a trigger signal generation unit 5a, an error signal storage unit 5b, an error signal comparison / determination unit 5c, and a step size setting unit 25d.
[0046]
The trigger signal generation unit 5a, error signal storage unit 5b, and error signal comparison / determination unit 5c receive the primary source Ps as an input to the trigger signal generation unit 5a, as in the first embodiment. A circuit unit that generates a pulse of the source Ps as a trigger pulse for signal processing performed by the error signal storage unit 5b and the error signal comparison / determination unit 5c. The error signal storage unit 5b is a RAM circuit, and the microphone 10 The error signal e (i) n input from is stored in a circuit portion for storing one period (one period as one data). Further, the error signal comparison / determination unit 5c includes the previous error signal (previous data) e (i) n-1 stored in the error signal storage unit 5b and the error signal input from the microphone 10 ( A circuit unit that obtains a difference (| e (i) n-1-e (i) n |) from e (i) n and compares it with a value α obtained and set in advance by experiments or the like. is there.
[0047]
Further, the step size setting unit 25d determines from the microphone 10 if | e (i) n-1-e (i) n | ≦ α based on the comparison result in the error signal comparison / determination unit 5c. It is determined that the input error signal e (i) n does not include noise components other than periodic sounds (random signals; human speech, music, sudden sounds, etc.), and the LMS arithmetic circuit 6 is informed. The step size μ is not changed. If | e (i) n-1-e (i) n |> α, | e (i) n in the current data e (i) n. The number of samples satisfying -1-e (i) n |> α is counted to obtain the ratio Sp. When this ratio is equal to or greater than the ratio β set in advance by experiments or the like, the LMS arithmetic circuit 6 The step size μ in the update equation for the filter coefficient W (n) of the adaptive filter 3 is smaller than And it has a circuit section for setting. Here, when Sp ≧ β, if the above ratio is Sp and the small step size is μ ′,
Sp = | e (i) n-1-e (i) n |> number of samples of α / e (i) total number of samples of n (2)
Also, the step size μ ′ is a value γ greater than 1, and
μ ′ = μ × (1 / γ) (3)
(However, it applies only to samples of | e (i) n-1-e (i) n |> α) | e (i) n-1-e (i) n |> α samples The update of the filter coefficient W (n) of the adaptive filter 3 in FIG.

It is done with the update formula.
[0048]
On the other hand, in the step size setting unit 25d, when the comparison result in the error signal comparison / determination unit 5c is | e (i) n-1-e (i) n |> α, and Sp <β Does not change the step size μ.
[0049]
Note that the step size μ ′ may be obtained by another formula, for example, using constants K and μ0,
μ ′ = K (1−Sp) + μ0
Or consider computer processing,
μ ′ = μ / (2) ^{(S p / β)} (3) ′
As described above, it may be continuously given by Sp 1.
[0050]
In the embodiment, Sp is the ratio of the number of samples of | e (i) n-1-e (i) n |> α as shown in the equation (2), but | e (i) n-1-e (I) The value of n | is the sum of all samples of e (i) n, ie,
Sp = Σ | e (i) n-1-e (i) n |
Sp = Σ (e (i) n-1-e (i) n) ² may be used.
[0051]
That is, in the second embodiment, the mixing of the noise component into the error signal is determined based on the ratio of the noise component mixing, and even if it is determined that the ratio of the noise component mixing is large, the adaptation speed is the same for all data. Instead of slowing down, only the sample with noise components is slowed down. For this reason, the adaptation to the periodic component to be silenced without the noise component does not deteriorate, and the noise component can be removed while keeping the silencing performance high.
[0052]
In addition, since the mixing of noise components into the error signal is determined based on the ratio of the mixing of the noise components, the adaptive speed is controlled according to the mixing amount of the noise components. However, it is possible to reliably prevent input.
[0062]
【The invention's effect】
As described above, according to the present invention, even if a random signal such as a human voice or music or a sudden sound is generated, the reverberation due to the random signal is suppressed without deteriorating the muffling performance. Is possible.
[Brief description of the drawings]
FIG. 1 is a system schematic diagram of a vehicle interior noise reduction apparatus according to a first embodiment of the present invention. FIG. 2 is an explanatory diagram of an input signal conversion circuit according to the first embodiment of the present invention. FIG. 4 is a system schematic diagram of a vehicle interior noise reduction device according to a second embodiment of the present invention. FIG. 5 is an explanatory diagram of an input signal conversion circuit according to the second embodiment of the present invention. Explanation of]
1 Engine 3 Adaptive filter (cancellation signal synthesis means)
5 Noise prevention circuit (noise prevention means)
6 LMS arithmetic circuit (filter coefficient updating means)
9 Speaker (offset sound generating means)
10 Microphone (error signal detection means)
25 Noise prevention circuit (noise prevention means)
e (i) n Error signal (error signal)
e (i) n-1 Previous data α Setting value β Setting ratio Ps Primary source (noise vibration source signal)
Sp | e (i) n-1−e (i) n |> ratio of the number of samples satisfying α> W (n) filter coefficient

Claims (2)

Cancel signal synthesizing means (3) for synthesizing a noise vibration source signal (Ps) highly correlated with engine vibration as a cancel signal by an adaptive filter;
Canceling sound generating means (9) for generating the cancel signal as a canceling sound for noise;
Error signal detection means (10) for detecting a noise reduction state at the listening point as an error signal (e (i) n);
At least one period of the error signal (e (i) n) is stored as one data, and the difference between the previous data (e (i) n-1) and the error signal (e (i) n) is calculated. When the difference is less than or equal to a preset value (α), the error signal (e (i) n) is output, while the previous data (e (i) n-1) and the error are output. When the difference from the signal (e (i) n) is larger than the set value (α), the previous data (e (i) n-1), the error signal (e (i) n) , A signal having the smallest absolute value is selected and output from a signal obtained by averaging the error signal (e (i) n) and the previous data (e (i) n-1) with a preset weight. Noise prevention means (5);
Filter coefficient updating means (6) for updating the filter coefficient (W (n)) of the adaptive filter (3) based on the noise vibration source signal (Ps) and the signal from the noise prevention means (5). Vehicle interior noise reduction device characterized by the above. Cancel signal synthesizing means (3) for synthesizing a noise vibration source signal (Ps) highly correlated with engine vibration as a cancel signal by an adaptive filter;
Canceling sound generating means (9) for generating the cancel signal as a canceling sound for noise;
Error signal detection means (10) for detecting a noise reduction state at the listening point as an error signal (e (i) n);
At least one period of the error signal (e (i) n) is stored as one data, and the difference between the previous data (e (i) n-1) and the error signal (e (i) n) is calculated. calculated, it is greater than the value the difference is preset (alpha), further, the previous data in the present error signal (e (i) n) ( e (i) n-1) and the current The number of samples whose difference from the error signal (e (i) n ) is larger than the set value (α) is obtained, and the ratio of the number of samples to the total number of samples of the current error signal (e (i) n ) (Sp ) And the difference between the previous data (e (i) n-1) and the error signal (e (i) n) when the ratio (Sp) is greater than or equal to a preset ratio (β). A noise prevention means (25) for setting a slow update rate of the filter coefficient (W (n)) of the adaptive filter (3) for a signal portion having a value greater than the set value (α);
Filter coefficient updating means (6) for updating the filter coefficient (W (n)) of the adaptive filter (3) based on the noise vibration source signal (Ps) and the error signal (e (i) n). Vehicle interior noise reduction device characterized by the above.

Images