JP3899023B2 - Hearing aid with adaptive filter to suppress acoustic feedback - Google Patents

Description

ここで、γは、ワーピング・パラメータである。これにより、ＦＩＲフィルタにおける固定遅延は、周波数に依存した遅延に置き換えられて、大きな遅延が低周波数で得られ、高周波数では遅延が小さくなる。また、全域通過要素(allpass elements)は、内部巡回形(internally recursive)であり、したがってワープ形ＦＩＲフィルタは無限インパルス応答を有することにも注意されたい。このように、ワープ形ＦＩＲという用語は、どちらかと言えば矛盾するが、トランスバーサル形(transversal)ＦＩＲフィルタとの構造的アナロジーをよく表現している。
【００５０】
この発明の実施例において、ワープ形ＦＩＲフィルタの次数は、同等の仕様のＦＩＲフィルタの次数より大幅に低くしうる。このため、所与の回路構成において、ワープ形ＦＩＲフィルタは、ＦＩＲフィルタより優れたフィルタ特性を提供しうる。さらに、ワーピング・パラメータγは、伝達関数、すなわち周波数スペクトルにおける共振および遮断周波数の位置決めを制御する制御パラメータとして用いることができ、これによってフィルタ出力信号と所望の信号との差であるエラー（誤差）信号ｅ(ｎ)のスペクトルを所望の周波数範囲内において最小にすることができる。
【００５１】
ＦＩＲフィルタまたはワープ形ＦＩＲフィルタにおいて、次のサンプルＹ(ｔ＋Ｔ)は、次式にしたがって計算される。
【数７】

である。
【００５２】
ｕは、信号ｕの最新のＮ個のサンプルを含むＮ次元ベクトルであり、ｃは、Ｎ次フィルタのＮ個の係数を含むベクトルであることに注意されたい。Ｔは、サンプリング周期である。
【００５３】
上式において、ｕ(ｔ）は、実時間（現在時間）ｔにおける実際値であり、ｕ(ｔ−ｉＴ）は、実時間ｔよりもｉ回サンプリング周期前における信号値である。離散時間系においては、簡略表記が用いられることがよくある。ここでは、記号ｕ(ｉ）は、時間ｔ−ｉＴにおける信号値を表す、すなわち上式におけるｕ(ｔ−ｉＴ)である。
【００５４】
たとえばクリュワー・アカデミック出版社から１９９７年に発行されたパウロＳ．Ｒ．ディニッツの適応フィルタリング（Adaptive Filtering by Paulo S. R. Diniz, Kluwer Academic Publishers, 1997）を参照すると、適応フィルタにおけるフィルタ係数の更新に下式の最小二乗平均(least mean square) アルゴリズムが用いられることはよく知られている。
【数８】

【００５５】
前記の簡略表記を用いると（ｎは実際のサンプルの参照番号である）、上式は次のように書き直される。
【数９】

【００５６】
または、さらに簡略化して、下記のように書き直される。
【数１０】

ここで、ｉは、個別のベクトル要素を示す。
【００５７】
フィルタ係数の更新には、次式の漏洩最小二乗平均(leaky least mean square)アルゴリズムを用いることが好ましい。
【数１１】

ここで、ｕｉは、ｎ 番目のサンプリング周期とｉ−１回の先行するサンプリング周期とにおいてディジタル処理装置の出力信号から導出される１組の信号値であり、ｃｉは１組のフィルタ係数であり、ｅはエラー信号の現在値であり、λ およびμは、スケーリング・ファクタである。μの値は、一般に、１０−６の大きさであり、λの値は、一般に約０．９９である。λは漏洩(leakage)(リーケージ）と呼ばれ、λ＜１のとき、フィルタ係数はそのそれぞれの初期値ｃｉ(０）の方にドリフトする。μは収束速度であり、適応フィルタが変化に適応する速度を決める。この適応速度は、μの値が増加するに伴って増加する。
【００５８】
さらに、アルゴリズムを正規化して、適応フィルタが入力信号の瞬間的な動的変化に実質的に応答しないようにすることが有利である。音響帰還信号を推定するためには、所望の入力信号は不適切であり、適応フィルタの収束性能を低下させる雑音になることに注意されたい。次式の正規化されたアルゴリズムは、正規化最小二乗平均(normalised Least Mean Square)（ｎＬＭＳ）アルゴリズムと呼ばれる。
【数１２】

【００５９】
しかし、上式において、パワーの計算には、多大な処理パワーが必要になり、このため、次式にしたがったパワー推定を用いることが好ましい。
【数１３】

ここで、αは、Ｐｕ推定が変化する速度を決定する所定の定数である。このアルゴリズムは、パワー正規化最小二乗平均アルゴリズムと呼ばれる。また、このパワー推定は、入力信号のパワーにおける突然の変化が適応アルゴリズムに及ぼす影響を最小にするように、入力トランスデューサからの出力信号に基づくことができる。
【００６０】
さらに、漏洩符号最小二乗平均(leaky sign least mean square)アルゴリズムと呼ばれる次式のような第３の更新アルゴリズムを用いて、適応フィルタ係数を更新してもよい。
【数１４】

ここで、μｓは、ｅ(ｎ)信号の符号(sign)とμとの積である。
【００６１】
さらにまた、漏洩符号−符号最小二乗平均(leaky sign-sign least mean square)アルゴリズムと呼ばれる次式のような第４の更新アルゴリズムを適応フィルタ係数に用いてもよい。
【数１５】

ここで、sgn（ｕｉ(ｎ)）は、ｕｉ(ｎ)の符号(sign)である。
【００６２】
フィルタ係数は、たとえば、他の信号と組み合わされるか、平均化されるか、そうでなければフィルタリングされる等の処理を受ける差分信号に基づいて更新してもよい。フィルタリングは、従来技術において知られている集束的態様(focussed manner)で行なわれうる。
【００６３】
また、この発明にしたがった多重チャネル（マルチチャネル）形補聴器においては、チャネルの適応フィルタは、同一個数のタップを有する必要はないことに注意されたい。たとえば、低周波チャネルにおいて動作する適応フィルタに、より多くのタップを、含ませることが望ましい。
【００６４】
上述のように、制御装置は、補聴器の音響帰還ループの第１のパラメータの決定に応じてλとμとを調節することができる。
【００６５】
補聴器のパラメータのさまざまな組(set)を、使用者が聴取することを望む、たとえばスピーチや音楽等の各種の音の種類のそれぞれに対して、および使用者が置かれる、たとえば静寂、騒音、反響、群集、屋外、室内、ヘッドセット等の各種の音響環境のそれぞれに対して設けることができる。たとえば、周波数の関数としての各種のゲイン設定を設けてもよく、入力信号レベルの関数としての各種のゲイン設定を設けてもよく、処理装置の動作ゲインの関数としての各種の収束速度を設けてもよいといった具合である。パラメータの各組は、補聴器の特定の動作モードを規定し、補聴器が特定のパラメータの組を用いて動作するときは、対応するモードで動作すると称せられる。したがって、特定の動作モードにおいては、補聴器の特定のパラメータ値が、特定の音響環境において対応する特定の音を適切に処理するために設定される。同様に、パラメータの自動調節が、現在の動作モードにしたがって行なわれうる。
【００６６】
音の種類（タイプ）は、使用者によって選択されるか、または、たとえば周波数分析、さまざまな周波数における信号対雑音比の分析、音響力学分析、言語認識、ニューラルネットワークによる認識等により、補聴器によって自動的に検出されうる。
【００６７】
同様に、音響環境の種類（タイプ）は、使用者によって選択されるか、または、たとえば周波数分析、さまざまな周波数における信号対雑音比の分析、音響力学分析、ニューラル・ネットワークによる認識等により、補聴器によって自動的に検出されうる。
【００６８】
たとえば、使用者が、音楽を聴取することを望む場合がある。第１の適応フィルタの第１の収束速度は、その場合には、音楽の自動相関に適合する値に設定されうる。さらに、ゲイン調節または第１の収束速度の調節もまた音楽の自動相関に適合して行なわれうる。第１の収束速度、たとえば１個以上のスケーリング・ファクタが、処理装置のゲインの関数として制御される場合は、その関数は、使用者が聴取すると決めた音楽やスピーチ等の特定の音を伴う特定の音響環境において用いるのに適合した関数の組の中から選択されうる。
【００６９】
さらに、調節（調整）もまた、たとえば音響帰還経路の帰還ゲイン等の、測定されるパラメータの変化速度にしたがって行なわれうる。
【００７０】
【実施例】
以下に、添付図面を参照してこの発明をさらに詳細に説明する。
【００７１】
図１は、この発明の実施例のブロック略図である。図１に示されている回路がディジタルもしくはアナログ回路またはこれらの組合せを用いて実現されうることは当業者に自明であろう。この実施例においては、ディジタル信号処理が採用され、したがって処理装置（プロセッサ）７および適応フィルタ１０はディジタル信号処理回路である。この実施例において、補聴器の全てのディジタル回路は、単一のディジタル信号処理チップ上に設けられるか、または複数個の集積回路チップ上に適切な方法で分散して設けられる。
【００７２】
この補聴器において、マイクロホン等の入力トランスデューサ１が用いられ、これによって、音声信号（音信号）が受信され、この音声信号が受信音声信号（受信音信号）を表す対応する電気信号に変換される。補聴器は複数個の入力トランスデューサ１を備えてもよく、これによってたとえばある一定方向の感知特性を提供できる。入力トランスデューサ１は伝達関数Ｈｍを有する。入力トランスデューサ１は、音声信号（音信号）をアナログ信号に変換する。このアナログ信号は、Ａ／Ｄ変換器（図示せず）によってサンプリングされるとともにディジタル化され、補聴器内においてディジタル信号処理されるディジタル信号４となる。ディジタル信号４は結合ノード９に送られ、この結合ノードにおいて、後述する帰還補償信号８５と組み合わせられる。結合ノード９は出力信号８６を出力し、この出力信号はディジタル信号処理装置（ディジタル信号プロセッサ）７に与えられ、所望の周波数特性と圧縮関数とにしたがって増幅されて、使用者の聴覚障害を補償するのに適した出力信号８０が得られる。
【００７３】
出力信号８０は、出力トランスデューサ５と任意の（オプションとしての）遅延Δとに送られ、遅延された信号８３が適応フィルタＡ、すなわち参照符号１０に供給される。出力トランスデューサ５は、出力信号８０を音響出力信号に変換する。この音響信号の一部分は、伝達関数Ｈｆｂをもつ帰還経路６に沿って入力トランスデューサ１に伝搬する。好ましくは、遅延線Δの時間遅延は、出力トランスデューサ５から入力トランスデューサ１までの帰還経路６に沿う通過時間と実質的に等しい。他の時間遅延を選択してもよい。しかし、より短い時間遅延または零の時間遅延は、フィルタリングを複雑にする。たとえばフィルタが有限インパルス応答フィルタである場合には、より長いフィルタ、すなわちより多くのタップを有するフィルタが必要になる。このため、さらに他の遅延を処理装置７の出力側において回路内に挿入して、遅延信号を出力トランスデューサ５とオプションの遅延Δとに供給し、これによって入力信号４とフィルタリングされた信号８５との間の相関を減少させてもよい。
【００７４】
適応フィルタ１０において、遅延信号８３がフィルタリングされて、音響帰還の推定であるフィルタリングされた信号８５が得られる。すなわちフィルタリングされた信号８５は、出力トランスデューサ５に由来する音を受信することによって生成されるトランスデューサ生成信号４の一部分の推定である。フィルタリングされた信号８５は、結合ノード９においてディジタル入力信号４から減じられ、これによって帰還補償信号８６が得られ、これがディジタル処理装置７に入力される。音響帰還経路における変化を補償するために、適応フィルタ１０のフィルタ係数は、フィルタリングされた信号８５が帰還経路６に沿って伝搬する帰還信号６と実質的に同一に維持されるように、連続的に更新される。
【００７５】
フィルタ１０は、上記したような漏洩符号−符号最小二乗平均（leaky sign-sign least mean square)アルゴリズムを有する有限インパルス応答（ＦＩＲ）フィルタまたはワープ形（warped）ＦＩＲフィルタである。
【００７６】
制御装置は、処理装置７における実際のゲインに応じてλとμとを調節（調整）する。ゲインの関数としてのスケーリング・ファクタ(scaling factors) λおよびμのグラフが図１０に示されている。これらの関数は、補聴器の動作モードに依存することに注意されたい。補聴器の現行（現在の）動作モードにしたがって制御装置１３により選択されうる、図１０に示されるような関数の選択可能なサブセットの組（セット）が提供される。さらにこれらの関数は、たとえば音響帰還経路内における減衰等の、測定されるパラメータの変化速度(rate)にしたがって選択される。
【００７７】
図１の実施例において、制御装置１３は、ディジタル処理装置７からライン１５を介して情報を受取る。ライン１５を介して受取られた、ディジタル処理装置７における現行動作ゲインに関する情報にしたがって、制御装置は、適応フィルタ１０のフィルタ係数の適応速度(adaptation rate)を調節(調整）する。この図において、破線および矢印は、処理信号の信号経路の一部分を構成しない制御ラインを示すことに注意されたい。
【００７８】
フィルタ１０の実施例であるＦＩＲフィルタが、図６にさらに詳細に示されている。簡単にするために、第１の４個のタップのみが図示されているが、フィルタは、任意の適切な個数のタップを含むことができる。演算子Ηを１に設定し、かつ演算子Β をμ(ｅ(ｎ)）に設定すると、漏洩最小二乗平均(leaky least mean square)アルゴリズムが達成（実現)される。λを１に設定すると、単純最小二乗平均(simple least mean square)アルゴリズムが達成される。Ηを１に設定し、かつΒをμsgn(ｅ(ｎ)）に設定すると、漏洩符号最小二乗平均アルゴリズム（leaky sign least square）が達成（実現）される。最後に、Η をsgn(ｕｉ（ｎ)）に設定し、かつΒをμsgn（ｅ(ｎ)）に設定して、漏洩符号−符号(leaky sign-sign)ＬＭＳアルゴリズムを達成（実現）することができる。また、フィルタ係数は、再帰的最小二乗(recursive least square)アルゴリズムを用いて計算してもよい。
【００７９】
フィルタ１０の実施例であるワープ形ＦＩＲフィルタは、図７にさらに詳細に示されている。図６および図７の上側の遅延線（ライン）より下の回路が同一であることに注意されたい。ワーピング(warping）パラメータγ は、0.5に等しいことが好ましい。γ＝０の場合、ワープ形ＦＩＲフィルタは、ＦＩＲフィルタに変わることに注意されたい。
【００８０】
図８に、ワープ形ＦＩＲフィルタの無限インパルス応答とＦＩＲフィルタの有限応答のグラフが示されている。このグラフから、ワープ形ＦＩＲフィルタでは、本来的に、ＦＩＲフィルタよりも、所望の伝達関数により近い値が得られることがわかる。
【００８１】
図９には、他のフィルタ１０４の所望の伝達関数Ｈに適応する適応フィルタ１０２の伝達関数 Ｈａを決定するテスト回路１００のブロック図が示されている。グラフの曲線により、適応フィルタ１０２がＦＩＲフィルタである場合のエラー（誤差）信号１０６のパワー・スペクトル１１０とともに、適応フィルタ１０２がワープ形ＦＩＲフィルタである場合のエラー（誤差）信号１０６のパワー・スペクトル１０８が示されている。ＦＩＲフィルタとワープ形ＦＩＲフィルタは、同一個数のタブを有する。６〜７kHz 未満において、ワープ形ＦＩＲフィルタは、エラー信号を１５dBまで改善することがわかる。出力トランスデューサ５の出力部が典型的には、６〜８kHz 程度の遮断周波数を有するため、８kHz を超える部分におけるワープ形ＦＩＲフィルタの性能は重要ではない。サンプリング周波数を変化させることによって、周波数軸に沿って示されている周波数値が変化することに注意されたい。また、γを調節（調整）してエラー信号１０６のスペクトルを、特定のタイプの聴覚障害等の特定の用途に合わせて最適化しうることにも注意されたい。
【００８２】
図２に、各チャネル（チャンネル）が図１に示された単一チャネルの実施例と一般に同じように動作する、この発明にしたがった補聴器の多重チャネル（マルチチャネル）の実施例が示されている。図２の参照符号に添え字が付け加えられていることを除いて、図１および図２の対応する部分は、同じ参照符号で示されている。簡単にするために、図２には３個のチャネルのみが図示されている。しかしながら、この補聴器は、チャネルをいかなる適切な個数でも含みうることに注意されたい。
【００８３】
図２にしたがったこの発明の多重チャネルの実施例は、図１に示された単一チャネルの実施例と同じ部分に加えて、帯域フィルタ信号（帯域フィルタリングされた信号）４ａ、４ｉ、４ｎを出力するフィルタ・バンク３を備えている。結合ノード９ａ、９ｉ、９ｎにおいて、それぞれの信号４ａ、４ｉ、４ｎが組み合わされて、それぞれの信号８６ａ、８６ｉ、８６ｎが形成される。これらの信号８６ａ、８６ｉ、８６ｎは、多重チャネル形ディジタル処理装置７に供給され、使用者の聴覚障害に合致する所望の特性にしたがって処理される。これは、個別のチャネルにおける異なるゲイン設定の調節を伴いうる。さらに、この処理には、圧縮機能(compressor functions) も含まれうる。さらにまた、雑音の減少等の他の機能が信号処理装置によって達成されうる。
【００８４】
ディジタル信号処理装置７からの出力信号はフィルタ・バンク１６に送られ、このフィルタ・バンクにおいて、１組の適応フィルタ１０ａ、１０ｉ、１０ｎにおける異なる周波数帯域またはチャネルに対応する帯域フィルタ信号（帯域フィルタリングされた信号）８３ａ、８３ｉ、８３ｎに分割される。好ましくは、フィルタ・バンク１６は、ディジタル四次フィルタからなる。
【００８５】
フィルタリングされた信号８５ａ、８５ｉ、８５ｎは、適応フィルタ１０ａ、１０ｉ、１０ｎから、それぞれの結合ノード９ａ、９ｉ、９ｎに供給されて、信号４ａ、４ｉ、４ｎから減じられ、信号８６ａ、８６ｉ、８６ｎが生成される。図１の実施例の場合と同様に、任意の遅延線Δにより出力信号８０を遅延させることができる。好ましくは、この遅延は、実質的に出力トランスデューサ５から入力トランスデューサ１までの音の最大伝搬時間に等しい。
【００８６】
処理装置７は、そのチャネルの信号を組み合わせて単一の出力信号８０にする。
【００８７】
多重チャネルの実施例においては、それぞれのチャネルの適応速度(adaptation rates)は、互いに異なりうる。このため、より高い適応速度を適用して、その結果として帰還共振が起こりやすい周波数で望ましくない歪みを発生させることができる。このことは、帰還共振が所望の信号にとって重要でない周波数で起こる場合に有利な特徴となる。
【００８８】
また、信号検出は、広範な周波数範囲において行なう方が困難である。このため、多重チャネルの装置では、単一チャネルの装置よりも、間違った信号検出による収束誤差(convergence errors)が生じにくい。
【００８９】
ひとつの実施例において、制御装置１３は、適応フィルタ１０、１０ａ、１０ｉ、１０ｎのフィルタ係数の適応速度を、Ｇ０からＧａのゲイン間隔において、処理装置における実際の動作ゲインの関数として制御する。
【００９０】
図３に図示されている補聴器は、図１の補聴器に追加の測定装置を設けたものに対応する。対応する部分は同一の参照符号で示されており、これらの部分の動作の説明を繰り返すことはしない。図３に示された補聴器は、さらに、第１の適応フィルタＡ，１０と並行して動作する、すなわち第１の適応フィルタＡ，１０と同じ信号に対して作用するが、第１の適応フィルタＡ，１０の第１の収束速度(convergence rate)より高速の第２の収束速度で動作する第２の適応フィルタＢ，参照符号１１を含む。第１の適応フィルタＡ，１０の出力８５は、結合ノード９に供給されて、信号４から減じられ、処理装置７に入力される信号８６が生成され、これによって音響帰還信号は、処理装置７による処理の前に、信号から実質的に除去される。第２の適応フィルタＢ，１１の出力８９は処理装置の入力の補正には用いられないことに注意されたい。
【００９１】
この実施例において、制御装置１３は、第２の適応フィルタＢ，１１のパラメータを決定することにより音響帰還の量を推定するようになっている。高い第１の収束速度により、第２の適応フィルタ１１は、第１の適応フィルタ１０よりも、音響帰還を経時的により厳密に追跡しうる。さらに、第２の適応フィルタ１１の出力信号８９が入力トランスデューサ信号４から減じられないため、所望の信号が第２の適応フィルタ１１によって歪まされることはない。
【００９２】
第１の適応フィルタ１０は、いかなる種類の適応フィルタであってもよいが、好ましくは、パワー正規化最小二乗平均(power-normalised Least Mean Square)（power−ｎＬＭＳ)アルゴリズムを用いるＦＩＲフィルタまたはワープ形ＦＩＲフィルタである。
【００９３】
第２の適応フィルタ１１は、フィルタリングされた信号８９を第２の結合ノード１２に出力し、このノードにおいて前記信号は第１の結合ノード９からの信号８６と組み合わされる。結合ノード１２からの出力信号９０は、第２の適応フィルタ１１に入力されて、フィルタ係数が調節される。
【００９４】
図３に示された実施例の重要な利点は、第２の適応フィルタ１１によって生成された出力信号が入力トランスデューサ１から出力トランスデューサ５までの主要信号経路に供給されないことである。この主要信号経路は、入力トランスデューサ１と、ディジタル変換手段（図示せず）と、結合ノード９と、ディジタル処理装置７と、出力トランスデューサ５とからなる。したがって、第２の適応フィルタ１１による信号処理が主要信号経路における信号に直接影響することはない。このため、主要信号経路における信号の信号歪みが第２の適応フィルタ１１によって創出されることはなく、したがって第２の適応フィルタ１１の適応速度（adaptation rate)を第１の適応フィルタ１０の適応速度より実質的に高くすることができる。第２の適応フィルタ１１の適応速度を第１の適応フィルタ１０の適応速度より実質的に高くすることができるため、帰還経路は、第２の適応フィルタ１１によって、第１の適応フィルタ１０によるよりも、経時的にはるかに厳密に監視されうる。好ましくは、第２の適応速度は、固定された高適応速度であるが、この適応速度は、たとえば１個またはそれ以上のスケーリング・ファクタを補正することによって調節されうる。たとえば、第２の適応フィルタの適応速度を、処理装置における実際のゲインまたは入力パワー・レベルにしたがって調節することが好ましいかもしれない。
【００９５】
適応速度(adaptation rate)の調節は、動作の異なるモード毎に相違しうる。
【００９６】
音響環境の急速な変化が起こった場合に、図３の第１の適応フィルタ１０は、こうした変化に即座に適応してこの変化を補償することができない。したがって、未補償の帰還信号が発生し始めることになる。しかしながら、第２の適応フィルタ１１は、第１の適応フィルタ１０よりはるかに高速であり、かつ帰還経路における変化に適応する。
【００９７】
ひとつの実施例において、制御装置は、帰還経路における変化に対する第２の適応フィルタ１１の迅速な応答に基づいて、たとえばμの値を制御するといったように、第１の適応フィルタ１０の適応速度を制御する。このため、第２の適応フィルタ１１の特性、たとえば減衰等のフィルタリング特性により帰還経路における変化が示されると、第１の適応フィルタ１０はそれに応じて制御される。たとえばゲインが帰還限界に近い場合は、第１の適応フィルタ１０の適応速度が増加する。第１の適応フィルタ１０の適応速度を増加させることにより、第１の適応フィルタは、より迅速に、たとえば音響帰還が望ましくない音の発生を招く前に、音響帰還の変化を補償しうる。
【００９８】
音響帰還の量は、好ましくは、第１の適応フィルタ１０のパラメータを決定すること、またはこれに代えて、もしくは加えて、第２の適応フィルタ１１のパラメータを決定することによって推定されうることに注意されたい。たとえば、それぞれの適応フィルタ１０、１１の入力信号と出力信号との比は、音響帰還経路を含む帰還経路の減衰の推定の要素となるので、この比を決定してもよい。さらに、このような計算を平均化された信号に基づいて行ない、雑音およびスピーチによる影響と収束誤差を抑制することが望ましい。これに代えて、所望の特性の平均を決定してもよい。好ましくは、上記のタイプのパワー推定を各信号に関して使用する。これに代えて、フィルタ係数を適切に変換することによって、適応フィルタ１０、１１のうちの一方のパラメータを決定してもよい。
【００９９】
また他の実施例において、制御装置は、帰還における変化が第２の適応フィルタ１１により検出されると、ディジタル処理装置におけるゲインを低下させる。特に、このことは、ディジタル処理装置の異なるチャネルにおいて選択的に実行されうる。
【０１００】
第１のパラメータの決定に基づいて、制御装置は、望ましくない音声（音）信号の発生を防ぐために処理装置が超えてはならない最大ゲイン値 Ｇｍａｘを計算することができる。多重チャネル形補聴器においては、各チャネル毎に別個のＧｍａｘ値があってもよい。
【０１０１】
さらに他の実施例において、制御装置は、Ｇ０からＧａへのゲイン間隔を変更する。このため、第２の適応フィルタ１１により、装置が不安定に近くなっていることが検出されると、この情報を用いてゲイン下限 Ｇ０を低下させ、これによって、 Ｇａを特定レベルに維持することが望まれる場合には、ゲイン間隔全体が下方に移動するか、またはゲイン間隔が拡大される。ゲイン下限 Ｇ０だけが変更される場合は、λおよびμの曲線が異なる間隔を包含するように変更されることが好ましい。
【０１０２】
これに関しては、ゲインとλおよびμとの関係は、図１０に示された関数とは異なることに注意されたい。
【０１０３】
図４に、各チャネルが一般に図３に示された単一チャネルの実施例と同じように動作する、この発明にしたがった補聴器の多重チャネル（マルチチャネル）の実施例が示されている。図３および図４の対応する部分は、図３の参照符号に添え字が付け加えられていることを除いて、同じ参照符号で示されている。簡単のために、３個のチャネルのみが図４に示されている。しかしながら、この補聴器は、同図に示されているようなチャネルを適切な個数で含みうることに注意されたい。簡単にするために、制御ラインは、図４においては省略されている。
【０１０４】
図４にしたがったこの発明の多重チャネルの実施例は、図３に示された単一チャネルの実施例と同じ構成部分に加えて、帯域フィルタ信号（帯域フィルタリングされた信号）８３ａ、８３ｉ、８３ｎを、第１の適応フィルタの組と、第２の適応フィルタ１１ａ、１１ｉ、１１ｎの組 (set)に出力するフィルタ・バンク１６を備えている。それぞれの第２の適応フィルタ１１ａ、１１ｉ、１１ｎの組は、それぞれの結合ノード１２ａ、１２ｉ、１２ｎに、結合ノード９ａ、９ｉ、９ｎからのそれぞれの信号８６ａ、８６ｉ、８６ｎと組み合わされるフィルタリングされた信号を供給する。
【０１０５】
図４に示された多重チャネルの実施例では、帰還経路の伝達関数のより詳細な推定が達成される。さらに、信号処理は、より低い周波数帯域においてより低いサンプリング周波数で行なわれ得、これはデシメーション(decimation)として知られた技術である。第２の適応フィルタの組からの出力信号は主要信号経路には供給されないため、いかなるアンチ・エリアジング(anti-aliasing) フィルタもシステムには必要とされないので、デシメーションは、これらのフィルタにおいて特に簡単に用いられうる。
【０１０６】
図４に示された実施例は、図３に示された実施例と同じように制御されうる。しかしながら、図４に示された実施例では、各個別のチャネルにおいてゲインを選択的に低下させることと、第１の適応フィルタ１０ａ、１０ｉ、１０ｎの組の各個別の適応フィルタの適応速度を選択的に調節することが可能である。このため、帰還共振が起こりにくい周波数においてゲインを高い値に維持することができ、かつ歪みを低レベルに維持することができるというさらに他の利点が得られる。
【０１０７】
図５に、図４に示された実施例と同様の構成であり、かつ同様に動作する多重チャネルの実施例が示されている。しかし、図５に示された実施例は、単一の適応フィルタ１０によって構成される第１の適応フィルタの組を有しており、かつ結合ノード９が単一の結合ノードであるため、より簡素である。
【０１０８】
多くのその他の実施例が、処理装置におけるチャネルと第１および第２の適応フィルタの組における適応フィルタの個数を変化させることによって得られる。また、処理装置におけるチャネルの個数は、第１の適応フィルタの組におけるフィルタの個数と異なっていてもよく、同様に前記フィルタの個数は、第２の適応フィルタの組におけるフィルタの個数と異なっていてもよい。
【０１０９】
特に、比較的少数のチャネルを有するディジタル信号処理装置７と、より多くのフィルタを含む第２の適応フィルタの組とを用いることが可能である。これに代え、第１の適応フィルタの組の個別の適応フィルタは、ディジタル信号処理装置７におけるチャネルとの組み合わせの上で動作するものでもよく、たとえばディジタル信号処理装置７の２個またはそれ以上のチャネルが第２の適応フィルタの組の特定の適応フィルタにより決定される同じ Ｇｍａｘを用いて動作するか、またはディジタル信号処理装置７の１個のチャネルが第２の適応フィルタの組の適応フィルタにより決定される２個以上のゲインのうちの最も低いゲインであるＧｍａｘ を用いて動作してもよい。しかしながら、現在のところは、単一の第１の適応フィルタ１０と多重チャネルの第２の適応フィルタ１１の組とを備えた実施例が好ましい。
【０１１０】
図１１に、周波数の関数としての動作ゲインのグラフが示されている。実線で示されている上側の曲線は、この発明にしたがった補聴器によって望ましくない音の発生を伴うことなしに得られる最大動作ゲインを示し、破線で示されている下側の曲線は、公知の補聴器の場合の対応するゲインを示す。
【図面の簡単な説明】
【図１】 この発明にしたがう補聴器のブロック図である。
【図２】 各チャネルが図１に示された補聴器に対応する多重チャネル形補聴器のブロック図である。
【図３】 この発明にしたがう測定装置を内蔵する補聴器のブロック図である。
【図４】 各チャネルが図３に示された補聴器に対応する多重チャネル形補聴器のブロック図である。
【図５】 単一の帯域適応フィルタを有する多重チャネル形補聴器のブロック図である。
【図６】 この発明にしたがう更新アルゴリズムを実現するＬＭＳ形ＦＩＲフィルタを示すブロック図である。
【図７】 この発明にしたがう更新アルゴリズムを実現するＬＭＳ形ワープＦＩＲフィルタを示すブロック図である。
【図８】 ＦＩＲフィルタのインパルス応答とワープ形ＦＩＲフィルタのインパルス応答との比較を示すグラフである。
【図９】 ＦＩＲフィルタおよびワープ形ＦＩＲフィルタの所望の伝達関数からの偏差を示すグラフである。
【図１０】 ディジタル処理装置におけるゲインに依存するフィルタ係数の変動を表す図である。
【図１１】 この発明により達成される最大可能ゲインの改善を示す図である。[0001]
【Technical field】
The present invention relates to a hearing aid including an adaptive filter that suppresses acoustic feedback in the hearing aid.
[0002]
[Prior art]
In hearing aid technology, it is well known that acoustic feedback leads to the generation of undesirable acoustic signals that can be heard by a hearing aid user.
[0003]
Acoustic feedback occurs when the hearing aid input transducer receives and detects the acoustic output signal generated by the output transducer. Amplification of the detected signal can lead to the generation of a stronger acoustic output signal, and the hearing aid may eventually oscillate.
[0004]
It is well known to provide an adaptive filter in a hearing aid to compensate for acoustic feedback. The adaptive filter estimates the transfer function from the output to the input of the hearing aid that includes the acoustic propagation path from the output transducer to the input transducer. The input of the adaptive filter is connected to the output of the hearing aid and the output signal of the adaptive filter is subtracted from the input transducer signal to compensate for acoustic feedback. This type of hearing aid is disclosed in US Pat. No. 5,402,496.
[0005]
In such a system, the adaptive filter acts to remove the correlation from the input signal, while the speech (language) and music signals are signals with significant autocorrelation. For this reason, removing correlations from signals representing speech and music will distort these signals, and such distortions are of course not desirable, so the adaptive filter cannot be adapted too quickly. Thus, the convergence rate of the adaptive filter in known hearing aids ensures that the desired high convergence rate that can handle sudden changes in the acoustic environment and the signals representing speech and music remain undistorted. A compromise with the desired low convergence rate that can be maintained.
[0006]
The lack of speed of adaptation still results in the generation of undesirable acoustic signals due to acoustic feedback. Unwanted acoustic signal generation is most likely at frequencies with high feedback loop gain. The loop gain is the product of the attenuation in the acoustic feedback path and the gain from the input to the output of the hearing aid.
[0007]
Acoustic feedback is achieved by using known CIC hearing aids (CIC = c omplete i n the c anal (in the complete auditory canal)). The short distance between the presence of the vent and the input / output transducer of the hearing aid leads to low attenuation in the acoustic feedback path from the output transducer to the input transducer, and the correlation between the signals is maintained due to the short delay time. It is.
[0008]
Various strategies are well known in the prior art to deal with acoustic feedback. For example, it is well known to keep the loop gain below a certain limit to prevent the occurrence of feedback resonance. Also known are phase adjustment (adjustment) of feedback signals, execution of frequency transpose, and compensation of feedback signals.
[0009]
Typically, the acoustic environment of a hearing aid changes over time and often changes rapidly over time in such a way that the propagation of sound from the hearing aid output transducer to its input transducer changes drastically. As an example, such a change can be caused by a change in the position of the user in the room, for example, from a free sound field position in the center of the room to a position near the wall that reflects the sound. Changes can also occur when the user yawns or when the user places the telephone handset on the ear. These various forms of changes are known to include changes of more than 20 dB in the attenuation of the feedback path, some of which may be almost instantaneous.
[0010]
It is known to keep the loop gain below the safe limit by limiting gain adjustment in the hearing aid to the maximum allowable gain based on experience. However, a large safety margin is required to deal with the above changes in the acoustic environment and the physical adaptation of the hearing aid to the wearer. It is also known to determine the maximum allowable gain when tailoring a hearing aid to a particular user. Nevertheless, a large safety margin is still necessary. In situations where the gain can be adjusted to a value higher than the maximum allowable gain without undesired sound generation, providing the above safety margins should make full use of the hearing aid capabilities. Hinder.
[0011]
In order to compensate for severe hearing impairment, it is desirable to be able to set a high gain in the hearing aid. However, the maximum gain that can be used is limited even in situations where attenuation in the acoustic feedback path is high due to the risk of oscillation, also called feedback resonance.
[0012]
German Patent Publication No. 19802568A and US Pat. No. 5,016,280 disclose hearing aids that include a measurement system for determining the characteristics of an acoustic return path. A test signal is transmitted through the system to determine the characteristics of the feedback path.
[0013]
In DE 19802568A, the coefficients of the digital filter are determined based on the impulse response of the feedback path, and in US Pat. No. 5,016,280, the filter coefficients of the adaptive compensation filter are transmitted in the feedback path. Calculated using a leaky LMS algorithm acting on the white noise signal.
[0014]
Each of these measuring devices is quite complex and the time for determination is relatively long, during which the normal functioning of the hearing aid is interrupted. For this reason, the determination is made only at certain occasions, for example when the user switches on the hearing aid. Therefore, a relatively high safety margin for gain is still needed to deal with changes in the acoustic environment between decision points.
[0015]
U.S. Pat. No. 5,619,580 discloses a hearing aid with an adaptive filter and a continuous working measurement device. A pseudo-random noise signal is added to the output signal. The monitoring system controls the hearing aid gain so that the loop gain is kept below a frequency dependent constant value. Although the filter coefficients of the adaptive filter are monitored and the update frequency of these coefficients is adjusted according to statistical analysis, this complicates the system. Another drawback of this system is that a noise generator is required and that the generated noise signal is always present. In addition, this system increases the adaptation rate. Therefore, even when the hearing aid is not operating close to resonance, the signal quality is degraded when a change in the acoustic environment is detected.
[0016]
DISCLOSURE OF THE INVENTION
For this reason, the hearing aid is improved so that the above-mentioned drawbacks can be overcome and the condition of safety margin of gain can be substantially eliminated, and the operating gain in a certain acoustic environment can be higher than that of a known hearing aid. It is needed.
[0017]
According to a first aspect of the present invention, the above and other objects are a method for suppressing acoustic feedback in a hearing aid, the step of converting an acoustic input signal into a first electrical signal, and the first electrical Splitting the signal into a first set of band-filtered first (multiple) electrical signals; processing each of the first band-filtered first electrical signals individually; and the processed electrical signals Are added to form a second electrical signal, the second electrical signal is converted to an acoustic output signal, and the second electrical signal is a band-filtered second (multiple) electrical signal. Dividing the second band-filtered electrical signal adaptively into the set and passing the filtered signal to the input side of the processor Estimating the acoustic feedback by generating a third signal by adapting to each signal at a respective first convergence rate, and determining a first parameter of the acoustic feedback loop of the hearing aid And compensating for acoustic feedback by adjusting (adjusting) the second parameter of the hearing aid in response to the first parameter, thereby reducing unwanted sound generation such as howling and signal distortion. It is a substantial prevention.
[0018]
According to a second aspect of the present invention, the above and other objects are achieved by a hearing aid with an adaptive filter that compensates for acoustic feedback. This adaptive filter serves to estimate the transfer function from the output part to the input part of the hearing aid including the acoustic propagation path from the output transducer to the input transducer. The input of the adaptive filter is connected to the electrical output of the hearing aid, and the output signal of the adaptive filter can be subtracted from the input transducer signal to compensate for acoustic feedback. The hearing aid further divides the first electrical signal into a set of band-filtered first (multiple) electrical signals (multiple (multiple)) and an input transducer that converts the acoustic input signal into a first electrical signal. ) By first processing each of the first filter bank having a bandpass filter and the bandpass filtered first electrical signal and adding the processed electrical signals to a second electrical signal; And a processing device for generating two electrical signals, and an output transducer for converting the second electrical signal into an acoustic output signal. The hearing aid further includes a second filter bank having (multiple) band-pass filters that divide the second electrical signal into a second band-filtered set of electrical signals (multiple). Generating a third (multiple) electrical signal by filtering the band-filtered electrical signal and each of the third electrical signal to a respective signal at the input side of the processing device. And a first set of adaptive filters having a first filter coefficient that estimates acoustic feedback by adapting at a convergence rate.
[0019]
A unique feature of this hearing aid is that the hearing aid determines the first parameter of the hearing aid's acoustic feedback loop and adjusts (adjusts) the second parameter of the hearing aid according to the first parameter. And a control device which substantially compensates for the generation of undesirable sounds.
[0020]
An important advantage of the present invention is that the controller automatically adjusts the parameters of the electronic feedback loop whenever the hearing aid operates at a high risk of unwanted sounds, and the generation of such sounds. Therefore, the condition of the gain safety margin is significantly relaxed.
[0021]
In the following, the frequency range of the bandpass filter is also called a channel.
[0022]
In a simple embodiment of the invention, the hearing aid is a single channel hearing aid, ie a hearing aid that processes incoming signals in only one frequency band. For this reason, the first filter bank is constituted by a single bandpass filter, which can be constituted by a bandpass filter that is unique in the electronic circuit, i.e. any special circuit is said bandpass filter. It does n’t mean that. Correspondingly, the addition of processed electrical signals in the processing device (processor) is reduced to the task of supplying a single processed electrical signal at the output of the processing device. Furthermore, the second filter bank is constituted by a single bandpass filter, and the first adaptive filter set is constituted by a single adaptive filter.
[0023]
In general, hearing impairment varies as a function of frequency in a different manner for each individual user. For this reason, the processing device is preferably divided into a plurality of channels so that each frequency band can receive different processing, for example, amplification using different gains. Correspondingly, the hearing aid can comprise a first set of adaptive filters having a plurality of adaptive filters that individually filter signals in respective frequency bands, thereby providing acoustic feedback in each channel of the hearing aid. Ability to control individually. Preferably, the frequency band of the first set of adaptive filters is substantially the same as the frequency band of the first filter bank so that the bandpass filter does not degrade the function of the adaptive filter.
[0024]
In one embodiment of the invention, the first set of adaptive filters subtracts the hearing aid electrical output from the input to the processor, and the difference signal is used to modify the filter coefficients, as described below. This difference signal is not used to correct the input signal to the processing device, thus avoiding signal distortion. Thus, in this embodiment of the invention, an acoustic feedback signal is estimated using the first adaptive filter without distorting the processed signal. Further, in this embodiment, at least one adaptive filter of the first set of adaptive filters operates on each decimated second band-filtered electrical signal so that the adaptive filter output signal is processed. Signal processing power requirements can be minimized without the need for additional filters.
[0025]
In another embodiment of the invention, the first set of adaptive filters subtracts the electrical output of the hearing aid from the electrical signal from the input transducer, and the difference signal is used to modify the filter coefficients and the input of the processor. So that the acoustic feedback signal is substantially removed from the signal prior to processing by the processing unit. In this embodiment, if a third filter bank is substantially identical to the first filter bank, the processed individual signal from each processor channel is output to the output signal from the processor. If added to the processing unit before summing to, signal decimation can be used in the processing unit and first adaptive filter set.
[0026]
The loop gain of the acoustic feedback loop, i.e. the gain of the acoustic feedback path from the output transducer to the input transducer, including the transducer transfer function, and the gain of the electronic circuit in the signal path from the input to the output of the hearing aid Can be avoided by monitoring the loop gain of the acoustic feedback loop, which is the sum of. When this loop gain approaches unity, a predetermined action is taken to prevent the generation of undesirable sounds. Since the first set of adaptive filters generates a signal corresponding to the signal generated by the acoustic feedback, by monitoring the attenuation in the first set of adaptive filters and the gain in the corresponding channel of the processing unit, the acoustic filter An indication of the loop gain of the feedback loop is obtained. Thus, for example, by determining the individual ratio between the amplitude of the signal at the input of the individual filter and the amplitude of the signal at the corresponding output of the individual filter, the controller can attenuate the first adaptive filter set. Can be configured to monitor. In addition, control can be achieved, for example, by similarly determining the input and output signal levels of individual processing unit (processor) channels, or by reading the value of a register in the processing unit including the current gain value of the individual processing unit channel The device may be configured to monitor the gain of individual channels of the processing device. In general, the gain of the processor channel is different for different channels and depends on the input level.
[0027]
Based on monitoring a first parameter of the acoustic feedback loop, such as loop gain, processor channel gain, attenuation of one adaptive filter of the first adaptive filter set, etc. Parameters can be adjusted to prevent unwanted sound generation. For example, the gain of at least one processing unit channel can be modified, eg, lowered, to keep the acoustic feedback loop gain below 1.
[0028]
The second parameter may be the maximum gain limit Gmax that the gain of the processing device must not exceed within a particular channel. The adaptation rate of the first set of adaptive filters is kept constant, while the maximum gain limit Gmax for a particular channel of the processing device is caused by a sudden change in the hearing aid in that channel, for example in the acoustic environment. It is always reduced when approaching a high risk of unwanted sound. For example, the maximum gain limit Gmax for a particular channel is reduced, while the first adaptive filter adapts to the changed acoustic environment and returns to its original value when the adaptive filter has finished adapting to the new situation. . Thus, no distortion of the desired signal occurs.
[0029]
An important advantage of this embodiment of the present invention is that the gain automatically drops as the feedback loop approaches resonance, so that the operating gain of the hearing aid is very high without the risk of unwanted sound. There is something that can be done. For this reason, the gain safety margin is not substantially required.
[0030]
In embodiments where the band filter (s) of the second filter bank are substantially the same as the respective band filters of the first filter bank, each channel is individually controlled based on decisions in that channel. Therefore, it is possible to prevent the gain from being lowered due to the influence of the frequency outside the channel.
[0031]
Furthermore, in an embodiment of the invention in which the differential signal from the first adaptive filter is supplied to the input of the processing device, the second parameter is the first convergence rate or adaptation rate of the first set of adaptive filters. It can be said. For example, the adaptation rate of the filter is such that whenever the hearing aid approaches a state where there is a high risk of unwanted sound, for example caused by a sudden change in the acoustic environment. It can depend on the operating gain of the processing device in such a way that the adaptation speed is increased and the change is compensated quickly.
[0032]
The convergence rate of the first set of adaptive filters can be adjusted by modifying the algorithm that updates the filter coefficients of the adaptive filter. As described further below, the algorithm can include one or more scaling facters that can be adjusted in response to the determination of the first parameter. For example, the one or more scaling factors may be adjusted as a predetermined function of the operating gain of the processing device.
[0033]
An important advantage of this embodiment is that the adaptive filter adapts more quickly to the situation as the acoustic feedback loop gain approaches resonance, thus reducing the operating gain of the hearing aid without the risk of unwanted sound. It can be very expensive. As already explained, the rapid adaptation of the adaptive filter causes the desired signal distortion. However, as soon as the adaptive filter finishes adapting, the convergence rate decreases and the desired signal is no longer distorted. Furthermore, this distortion can occur in a frequency band that does not affect the understanding of the received voice (sound) signal.
[0034]
A gain interval from G0 to Ga can be provided in the hearing aid. G0 is a predetermined lower limit of gain below which no feedback resonance and undesirable sound can occur. G0 can be determined by a mounting adjustment procedure. Ga is an adjustable (adjustable) upper limit of gain that is adjusted (adjusted) according to the desired voice quality. Preferably, Ga is adjusted by a procedure for mounting adjustment.
[0035]
The convergence speed can vary as a predetermined function, such as a linear or non-linear function, of the gain of the processing device, for example within the range of G0 to Ga. For example, one or more scaling factors of the adaptive filter update algorithm may vary as a predetermined function, such as a linear or non-linear function of the gain of the processor, for example, in the range of G0 to Ga.
[0036]
When adjusting the hearing aid for each user (user), the transfer characteristic of the feedback path is measured. Based on these characteristics, G0 and Ga values with appropriate safety margins are determined and stored in the hearing aid. Several factors are taken into account to determine G0. The feedback path characteristic is not constant as already described. For this reason, if feedback compensation is too slow, sudden changes can lead to feedback resonance. In addition, it may be difficult to predict the magnitude and duration of the change in attenuation of the feedback path. On the other hand, rapid adaptation can lead to unacceptable distortion of the desired signal, and the level of unacceptable distortion is also a subjective magnitude.
[0037]
However, in a situation where the characteristics of the acoustic feedback path have been stable for a certain period, it is possible to accurately estimate the characteristics of the feedback path. This is because in these situations, the relationship between the signal at the input of the first set of adaptive filters and the signal at the output of the first set of adaptive filters is such that the characteristics of the acoustic feedback path, eg attenuation, are accurate. This is because it becomes an important measure. Knowing the gain characteristics of the digital processor and the gain characteristics of the acoustic feedback signal, an estimate for the acoustic feedback loop can be obtained. With this knowledge, dynamically changing G0 values can be introduced into hearing aids. In one embodiment, the interval from G0 to Ga has a fixed magnitude, regardless of the change in G0, ie the entire interval can move according to the change in G0.
[0038]
According to a preferred embodiment of the invention, the hearing aid operates in parallel with the first set of adaptive filters, i.e. acts on the same signal, but the first convergence of the first set of adaptive filters. A second set of adaptive filters having a second convergence rate that is faster than the velocity is further provided. The output of the first adaptive filter set is fed to a corresponding input of the processing device, so that the acoustic feedback signal is substantially removed from the signal prior to processing by the processing device. The output of the second set of adaptive filters is not used to modify the input signal of the processing device.
[0039]
In this embodiment, the controller is configured to estimate the amount of acoustic feedback by determining the parameters of the second adaptive filter set. Due to the fast second convergence rate, the second adaptive filter can track acoustic feedback more closely over time than the first adaptive filter. Further, since the output signal of the second adaptive filter is not subtracted from the input transducer signal, the desired signal is not distorted by the second adaptive filter.
[0040]
Thus, according to a preferred embodiment of the present invention, the second filter coefficient for suppressing feedback in the hearing aid by filtering the second band-filtered electric signal into each fourth electric signal is provided. Generating a fifth electrical signal by subtracting the fourth electrical signal from the respective first band-filtered electrical signal and processing the fifth electrical signal; There is provided a hearing aid further comprising a coupling node for supplying to the device, wherein the second filter coefficient is updated at a second convergence rate that is faster than the first convergence rate.
[0041]
The amount of acoustic feedback can be estimated by determining the respective ratios of the signal amplitude at the input of the second adaptive filter set and the signal amplitude at the output of the second adaptive filter set. . This method achieves a quick response to changes in the acoustic return path and requires very little processor power.
[0042]
The second parameter may be a second convergence rate or adaptation rate of the first set of adaptive filters. For example, the adaptive speed of filtering increases the adaptive speed of the first adaptive filter whenever the hearing aid approaches a state where there is a high risk of unwanted sound, for example caused by a sudden change in the acoustic environment. Thus, it may depend on the operating gain of the processing device or the attenuation of the second set of adaptive filters, or a combination thereof, in such a way as to quickly compensate for this change.
[0043]
As already described with respect to the second set of adaptive filters, the convergence rate of the first set of adaptive filters can be adjusted by modifying an algorithm that updates the filter coefficients of the adaptive filter. As described further below, the algorithm can include one or more scaling factors that can be adjusted in response to the determination of the second parameter. For example, one or more scaling factors can be set as a predetermined function of the operating gain of the processing device.
[0044]
The first set of adaptive filters provides individual filtering of the signal in each frequency band. Preferably, the frequency band of the first set of adaptive filters is substantially the same as the frequency band of the second filter bank.
[0045]
The frequency bands of the first adaptive filter set may differ in number and range from the frequency bands of the first filter bank and the second adaptive filter set. However, in the preferred embodiment of the invention, the second filter bank comprises a plurality of bandpass filters, while the first set of adaptive filters is a modification of the processor input signal in a single frequency band. Is constituted by a single adaptive filter. Therefore, a simple single-band acoustic feedback compensation loop can be provided in a hearing aid having frequency dependent compensation capability.
[0046]
Thus, according to a preferred embodiment of the present invention, a second adaptation having a second filter coefficient that suppresses feedback in the hearing aid by filtering the second electrical signal into a fourth electrical signal. A fifth electrical signal is generated by subtracting the fourth electrical signal from the filter and the first electrical signal, and the fifth electrical signal is supplied to each bandpass filter of the first filter bank. A hearing aid is further provided, wherein the second filter coefficients are updated at a second convergence rate that is faster than the first convergence rate.
[0047]
Thus, in a preferred embodiment of the present invention, the processing unit and the second adaptive filter are divided into channel (s) covering the same frequency band (s), while the first adaptive filter is It is not divided into a plurality of channels. Furthermore, the control device may be configured to control the individual maximum gain limit Gmax of each processing device (processor) channel in response to determining the attenuation of the channel of the corresponding first adaptive filter. The controller may further reduce the second convergence rate of one filter of the second set of adaptive filters when the gain of the corresponding processor channel is limited by the Gmax limit, thereby reducing the period of gain limitation. You may comprise so that it may raise. Further, the controller may be configured to adjust the gain limit and / or convergence speed according to the current operating mode of the hearing aid. The term operating mode is described below.
[0048]
Preferably, the at least one adaptive filter is a finite impulse response (FIR) filter, and more preferably, the at least one adaptive filter is a warped FIR filter or a warped infinite impulse response (infinite impulse response). ) A warped filter such as an (IIR) filter.
[0049]
In the warped FIR filter of this example, unit delays are replaced by first order allpass sections. However, warping can also be realized using secondary or higher order all-pass sections. The primary all-pass section has the following z-transform.
[Formula 6]

Here, γ is a warping parameter. Thereby, the fixed delay in the FIR filter is replaced with a delay depending on the frequency, and a large delay is obtained at a low frequency, and the delay becomes small at a high frequency. It should also be noted that allpass elements are internally recursive and therefore warped FIR filters have an infinite impulse response. Thus, the term warp FIR is somewhat contradictory, but well represents a structural analogy with a transversal FIR filter.
[0050]
In an embodiment of the invention, the order of the warp FIR filter can be significantly lower than the order of an equivalent FIR filter. For this reason, in a given circuit configuration, the warped FIR filter can provide better filter characteristics than the FIR filter. Further, the warping parameter γ can be used as a transfer function, ie, a control parameter that controls the positioning of the resonance and cut-off frequency in the frequency spectrum, thereby an error that is the difference between the filter output signal and the desired signal. The spectrum of the signal e (n) can be minimized within the desired frequency range.
[0051]
In the FIR filter or the warped FIR filter, the next sample Y (t + T) is calculated according to the following equation.
[Expression 7]

It is.
[0052]
u Is an N-dimensional vector containing the latest N samples of the signal u, c Note that is a vector containing N coefficients of an Nth order filter. T is a sampling period.
[0053]
In the above equation, u (t) is an actual value at real time (current time) t, and u (t−iT) is a signal value i sampling periods before the actual time t. In discrete time systems, the simplified notation is often used. Here, the symbol u (i) represents a signal value at time t−iT, that is, u (t−iT) in the above equation.
[0054]
For example, Paul S. A. published in 1997 by Krewer Academic Publishers. R. Referring to Dinitz's adaptive filtering (Adaptive Filtering by Paulo SR Diniz, Kluwer Academic Publishers, 1997), it is well known that the least mean square algorithm is used to update filter coefficients in adaptive filters: ing.
[Equation 8]

[0055]
Using the above shorthand notation (n is the actual sample reference number), the above equation is rewritten as:
[Equation 9]

[0056]
Or, it is further simplified and rewritten as follows.
[Expression 10]

Here, i represents an individual vector element.
[0057]
For updating the filter coefficients, it is preferable to use a leaky least mean square algorithm:
[Expression 11]

Here, ui is a set of signal values derived from the output signal of the digital processing device in the nth sampling period and i-1 preceding sampling periods, and ci is a set of filter coefficients. , E is the current value of the error signal and λ 1 and μ are scaling factors. The value of μ is typically on the order of 10 −6 and the value of λ is generally about 0.99. λ is called leakage (leakage), and when λ <1, the filter coefficients drift towards their respective initial values ci (0). μ is the convergence speed and determines the speed at which the adaptive filter adapts to changes. This adaptation speed increases as the value of μ increases.
[0058]
Furthermore, it is advantageous to normalize the algorithm so that the adaptive filter does not substantially respond to instantaneous dynamic changes in the input signal. Note that the desired input signal is inappropriate for estimating the acoustic feedback signal, resulting in noise that degrades the convergence performance of the adaptive filter. The normalized algorithm of the following equation is called the normalized Least Mean Square (nLMS) algorithm.
[Expression 12]

[0059]
However, in the above equation, a large amount of processing power is required for the calculation of power. For this reason, it is preferable to use power estimation according to the following equation.
[Formula 13]

Here, α is a predetermined constant that determines the speed at which the Pu estimation changes. This algorithm is called the power normalized least mean square algorithm. This power estimation can also be based on the output signal from the input transducer so as to minimize the impact of sudden changes in the power of the input signal on the adaptive algorithm.
[0060]
Further, the adaptive filter coefficient may be updated using a third update algorithm such as the following equation called a leaky sign least mean square algorithm.
[Expression 14]

Here, μs is a product of the sign of the e (n) signal and μ.
[0061]
Furthermore, a fourth update algorithm called a leaky code-sign least mean square algorithm such as the following equation may be used for the adaptive filter coefficient.
[Expression 15]

Here, sgn (ui (n)) is a sign of ui (n).
[0062]
The filter coefficients may be updated based on differential signals that are subject to processing, eg, combined with other signals, averaged, or otherwise filtered. Filtering can be done in a focused manner as known in the prior art.
[0063]
It should also be noted that in a multi-channel hearing aid according to the present invention, the channel adaptive filters need not have the same number of taps. For example, it may be desirable to include more taps in an adaptive filter that operates in a low frequency channel.
[0064]
As described above, the control device can adjust λ and μ in response to the determination of the first parameter of the acoustic feedback loop of the hearing aid.
[0065]
Different sets of hearing aid parameters are desired for the user to listen to, for each of the various sound types, e.g. speech and music, and where the user is placed, e.g. silence, noise, It can be provided for each of various acoustic environments such as reverberation, crowds, outdoors, indoors, and headsets. For example, various gain settings as a function of frequency may be provided, various gain settings as a function of input signal level may be provided, and various convergence speeds may be provided as a function of processor operating gain. It is a good condition. Each set of parameters defines a specific mode of operation of the hearing aid, and when the hearing aid operates with a specific set of parameters, it is said to operate in the corresponding mode. Thus, in a specific mode of operation, a specific parameter value of the hearing aid is set to properly process a corresponding specific sound in a specific acoustic environment. Similarly, automatic parameter adjustment can be made according to the current operating mode.
[0066]
The type of sound is selected by the user or automatically by the hearing aid, eg by frequency analysis, signal-to-noise ratio analysis at various frequencies, acoustic mechanics analysis, language recognition, neural network recognition, etc. Can be detected automatically.
[0067]
Similarly, the type of acoustic environment is selected by the user, or the hearing aid, for example, by frequency analysis, signal-to-noise analysis at various frequencies, acoustic mechanics analysis, neural network recognition, etc. Can be detected automatically.
[0068]
For example, a user may wish to listen to music. The first convergence speed of the first adaptive filter can then be set to a value that matches the automatic correlation of the music. Further, gain adjustment or first convergence speed adjustment may also be made in conformity with the autocorrelation of music. If the first convergence speed, eg one or more scaling factors, is controlled as a function of the gain of the processing device, that function is accompanied by a specific sound such as music or speech that the user has decided to listen to. It can be selected from a set of functions that are adapted for use in a particular acoustic environment.
[0069]
Furthermore, adjustment (adjustment) can also be made according to the rate of change of the parameter being measured, for example the feedback gain of the acoustic feedback path.
[0070]
【Example】
Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings.
[0071]
FIG. 1 is a schematic block diagram of an embodiment of the present invention. Those skilled in the art will appreciate that the circuit shown in FIG. 1 can be implemented using digital or analog circuits or combinations thereof. In this embodiment, digital signal processing is employed, so that the processing unit (processor) 7 and the adaptive filter 10 are digital signal processing circuits. In this embodiment, all the digital circuits of the hearing aid are provided on a single digital signal processing chip or distributed in a suitable manner on a plurality of integrated circuit chips.
[0072]
In this hearing aid, an input transducer 1 such as a microphone is used, whereby an audio signal (sound signal) is received, and this audio signal is converted into a corresponding electrical signal representing a received audio signal (received sound signal). The hearing aid may comprise a plurality of input transducers 1, which can provide, for example, a certain direction of sensing characteristics. The input transducer 1 has a transfer function Hm. The input transducer 1 converts an audio signal (sound signal) into an analog signal. This analog signal is sampled and digitized by an A / D converter (not shown) to become a digital signal 4 which is digital signal processed in the hearing aid. The digital signal 4 is sent to a coupling node 9 where it is combined with a feedback compensation signal 85 which will be described later. The coupling node 9 outputs an output signal 86 which is fed to a digital signal processor (digital signal processor) 7 and amplified according to the desired frequency characteristics and compression function to compensate for the hearing impairment of the user. An output signal 80 suitable for this is obtained.
[0073]
The output signal 80 is sent to the output transducer 5 and an optional (optional) delay Δ, and the delayed signal 83 is fed to the adaptive filter A, ie reference numeral 10. The output transducer 5 converts the output signal 80 into an acoustic output signal. A portion of this acoustic signal propagates to the input transducer 1 along a feedback path 6 having a transfer function Hfb. Preferably, the time delay of the delay line Δ is substantially equal to the transit time along the feedback path 6 from the output transducer 5 to the input transducer 1. Other time delays may be selected. However, shorter time delays or zero time delays complicate filtering. For example, if the filter is a finite impulse response filter, a longer filter, i.e. a filter with more taps, is required. For this purpose, further delays are inserted into the circuit at the output side of the processing unit 7 to supply the delayed signal to the output transducer 5 and the optional delay Δ, whereby the input signal 4 and the filtered signal 85 The correlation between may be reduced.
[0074]
In the adaptive filter 10, the delayed signal 83 is filtered to obtain a filtered signal 85 that is an estimate of acoustic feedback. That is, the filtered signal 85 is an estimate of a portion of the transducer generation signal 4 that is generated by receiving sound originating from the output transducer 5. The filtered signal 85 is subtracted from the digital input signal 4 at the coupling node 9, thereby obtaining a feedback compensation signal 86, which is input to the digital processor 7. To compensate for changes in the acoustic feedback path, the filter coefficients of the adaptive filter 10 are continuous so that the filtered signal 85 remains substantially the same as the feedback signal 6 propagating along the feedback path 6. Updated to
[0075]
Filter 10 is a finite impulse response (FIR) filter or a warped FIR filter having a leaky sign-sign least mean square algorithm as described above.
[0076]
The control device adjusts (adjusts) λ and μ according to the actual gain in the processing device 7. A graph of scaling factors λ and μ as a function of gain is shown in FIG. Note that these functions depend on the operating mode of the hearing aid. A set of selectable subsets of functions as shown in FIG. 10 is provided that can be selected by the controller 13 according to the current (current) mode of operation of the hearing aid. Furthermore, these functions are selected according to the rate of change of the parameter to be measured, for example the attenuation in the acoustic feedback path.
[0077]
In the embodiment of FIG. 1, the control device 13 receives information from the digital processing device 7 via line 15. According to the information about the current operating gain in the digital processing device 7 received via the line 15, the control device adjusts (adjusts) the adaptation rate of the filter coefficients of the adaptive filter 10. Note that in this figure, dashed lines and arrows indicate control lines that do not form part of the signal path of the processed signal.
[0078]
An FIR filter that is an example of filter 10 is shown in more detail in FIG. For simplicity, only the first four taps are shown, but the filter can include any suitable number of taps. If the operator Η is set to 1 and the operator Β is set to μ (e (n)), a leaky least mean square algorithm is achieved (implemented). Setting λ to 1 achieves a simple least mean square algorithm. Setting Η to 1 and Β to μsgn (e (n)) achieves (realizes) a leaky sign least squares algorithm. Finally, achieve (realize) the leaky sign-sign LMS algorithm by setting Η to sgn (ui (n)) and Β to μsgn (e (n)). Can do. The filter coefficients may also be calculated using a recursive least square algorithm.
[0079]
A warp FIR filter, which is an example of filter 10, is shown in more detail in FIG. Note that the circuits below the upper delay line (line) in FIGS. 6 and 7 are identical. The warping parameter γ is preferably equal to 0.5. Note that when γ = 0, the warped FIR filter is changed to an FIR filter.
[0080]
FIG. 8 shows a graph of the infinite impulse response of the warped FIR filter and the finite response of the FIR filter. From this graph, it can be seen that the warp FIR filter inherently provides a value closer to the desired transfer function than the FIR filter.
[0081]
FIG. 9 shows a block diagram of a test circuit 100 that determines the transfer function Ha of the adaptive filter 102 that adapts to the desired transfer function H of another filter 104. The curve in the graph shows the power spectrum of the error signal 106 when the adaptive filter 102 is a FIR filter and the power spectrum of the error signal 106 when the adaptive filter 102 is a warped FIR filter. 108 is shown. The FIR filter and the warp FIR filter have the same number of tabs. It can be seen that at less than 6-7 kHz, the warped FIR filter improves the error signal to 15 dB. Since the output section of the output transducer 5 typically has a cut-off frequency on the order of 6-8 kHz, the performance of the warped FIR filter in the portion above 8 kHz is not important. Note that changing the sampling frequency changes the frequency values shown along the frequency axis. Note also that γ can be adjusted to optimize the spectrum of the error signal 106 for a particular application, such as a particular type of hearing impairment.
[0082]
FIG. 2 shows a multi-channel embodiment of a hearing aid according to the present invention in which each channel (channel) operates generally the same as the single-channel embodiment shown in FIG. Yes. Corresponding parts in FIGS. 1 and 2 are indicated with the same reference numerals, except that subscripts are added to the reference numerals in FIG. For simplicity, only three channels are shown in FIG. However, it should be noted that this hearing aid can include any suitable number of channels.
[0083]
The multi-channel embodiment of the present invention according to FIG. 2 adds band-filter signals (band-filtered signals) 4a, 4i, 4n in addition to the same parts as the single-channel embodiment shown in FIG. An output filter bank 3 is provided. At coupling nodes 9a, 9i, 9n, the respective signals 4a, 4i, 4n are combined to form respective signals 86a, 86i, 86n. These signals 86a, 86i, 86n are fed to the multichannel digital processor 7 and processed according to desired characteristics that match the hearing impairment of the user. This may involve adjusting different gain settings in individual channels. In addition, this process may include compressor functions. Furthermore, other functions such as noise reduction can be achieved by the signal processing device.
[0084]
The output signal from the digital signal processor 7 is sent to a filter bank 16 where band filter signals (band-filtered) corresponding to different frequency bands or channels in the set of adaptive filters 10a, 10i, 10n. Signal) 83a, 83i, 83n. Preferably, the filter bank 16 comprises a digital fourth order filter.
[0085]
The filtered signals 85a, 85i, 85n are supplied from the adaptive filters 10a, 10i, 10n to the respective coupling nodes 9a, 9i, 9n and subtracted from the signals 4a, 4i, 4n, and the signals 86a, 86i, 86n. Is generated. As in the case of the embodiment of FIG. 1, the output signal 80 can be delayed by an arbitrary delay line Δ. Preferably, this delay is substantially equal to the maximum sound propagation time from the output transducer 5 to the input transducer 1.
[0086]
The processing device 7 combines the signals of the channels into a single output signal 80.
[0087]
In multi-channel embodiments, the adaptation rates of each channel can be different from each other. Thus, higher adaptation speeds can be applied, resulting in undesirable distortion at frequencies where feedback resonance is likely to occur. This is an advantageous feature when the feedback resonance occurs at a frequency that is not critical to the desired signal.
[0088]
Moreover, it is more difficult to perform signal detection in a wide frequency range. Thus, convergence errors due to incorrect signal detection are less likely to occur in multi-channel devices than in single-channel devices.
[0089]
In one embodiment, the control device 13 controls the adaptive speed of the filter coefficients of the adaptive filters 10, 10a, 10i, 10n as a function of the actual operating gain in the processing device in the gain interval from G0 to Ga.
[0090]
The hearing aid shown in FIG. 3 corresponds to the hearing aid of FIG. 1 provided with an additional measuring device. Corresponding parts are denoted by the same reference numerals and the description of the operation of these parts will not be repeated. The hearing aid shown in FIG. 3 further operates in parallel with the first adaptive filter A, 10, ie acts on the same signal as the first adaptive filter A, 10, but the first adaptive filter A second adaptive filter B operating at a second convergence rate faster than the first convergence rate of A, 10 and a reference numeral 11 are included. The output 85 of the first adaptive filter A, 10 is supplied to the coupling node 9 and is subtracted from the signal 4 to generate a signal 86 that is input to the processing device 7, whereby the acoustic feedback signal is converted into the processing device 7. Is substantially removed from the signal prior to processing by. Note that the output 89 of the second adaptive filter B, 11 is not used to correct the input of the processor.
[0091]
In this embodiment, the controller 13 estimates the amount of acoustic feedback by determining the parameters of the second adaptive filters B and 11. Due to the high first convergence rate, the second adaptive filter 11 can track acoustic feedback more closely over time than the first adaptive filter 10. Furthermore, since the output signal 89 of the second adaptive filter 11 is not subtracted from the input transducer signal 4, the desired signal is not distorted by the second adaptive filter 11.
[0092]
The first adaptive filter 10 may be any type of adaptive filter, but is preferably an FIR filter or warped using a power-normalized Least Mean Square (power-nLMS) algorithm. FIR filter.
[0093]
The second adaptive filter 11 outputs a filtered signal 89 to the second combining node 12 where it is combined with the signal 86 from the first combining node 9. The output signal 90 from the combination node 12 is input to the second adaptive filter 11 and the filter coefficient is adjusted.
[0094]
An important advantage of the embodiment shown in FIG. 3 is that the output signal generated by the second adaptive filter 11 is not supplied to the main signal path from the input transducer 1 to the output transducer 5. This main signal path consists of an input transducer 1, digital conversion means (not shown), a coupling node 9, a digital processor 7 and an output transducer 5. Therefore, the signal processing by the second adaptive filter 11 does not directly affect the signal in the main signal path. For this reason, signal distortion of the signal in the main signal path is not created by the second adaptive filter 11, and therefore the adaptation rate of the second adaptive filter 11 is set to the adaptation rate of the first adaptive filter 10. It can be made substantially higher. Since the adaptive speed of the second adaptive filter 11 can be made substantially higher than the adaptive speed of the first adaptive filter 10, the feedback path is changed by the second adaptive filter 11 than by the first adaptive filter 10. Can also be monitored much more closely over time. Preferably, the second adaptation speed is a fixed high adaptation speed, but this adaptation speed can be adjusted, for example, by correcting one or more scaling factors. For example, it may be preferable to adjust the adaptation speed of the second adaptive filter according to the actual gain or input power level in the processing unit.
[0095]
Adjustment of the adaptation rate may be different for different modes of operation.
[0096]
When a rapid change in the acoustic environment occurs, the first adaptive filter 10 of FIG. 3 cannot immediately adapt to such a change and compensate for this change. Therefore, an uncompensated feedback signal starts to be generated. However, the second adaptive filter 11 is much faster than the first adaptive filter 10 and adapts to changes in the feedback path.
[0097]
In one embodiment, the controller sets the adaptive speed of the first adaptive filter 10 based on the quick response of the second adaptive filter 11 to changes in the feedback path, such as controlling the value of μ. Control. Therefore, when a change in the feedback path is indicated by a characteristic of the second adaptive filter 11, for example, a filtering characteristic such as attenuation, the first adaptive filter 10 is controlled accordingly. For example, when the gain is close to the feedback limit, the adaptive speed of the first adaptive filter 10 increases. By increasing the adaptation speed of the first adaptive filter 10, the first adaptive filter can compensate for changes in acoustic feedback more quickly, for example, before acoustic feedback results in undesirable sound generation.
[0098]
The amount of acoustic feedback can preferably be estimated by determining the parameters of the first adaptive filter 10, or alternatively or additionally, determining the parameters of the second adaptive filter 11. Please be careful. For example, since the ratio between the input signal and the output signal of each adaptive filter 10 and 11 is an element for estimating attenuation of the feedback path including the acoustic feedback path, this ratio may be determined. Furthermore, it is desirable to perform such calculations based on the averaged signal to suppress the effects of noise and speech and convergence errors. Alternatively, an average of desired characteristics may be determined. Preferably, the above type of power estimation is used for each signal. Alternatively, one parameter of the adaptive filters 10 and 11 may be determined by appropriately converting the filter coefficient.
[0099]
In another embodiment, the controller reduces the gain in the digital processing device when a change in feedback is detected by the second adaptive filter 11. In particular, this can be selectively performed on different channels of the digital processing device.
[0100]
Based on the determination of the first parameter, the control device can calculate a maximum gain value Gmax that the processing device must not exceed in order to prevent the generation of undesirable speech (sound) signals. In a multichannel hearing aid, there may be a separate Gmax value for each channel.
[0101]
In yet another embodiment, the control device changes the gain interval from G0 to Ga. For this reason, when the second adaptive filter 11 detects that the device is nearly unstable, this information is used to lower the gain lower limit G0, thereby maintaining Ga at a specific level. Is desired, the entire gain interval is moved downward or the gain interval is increased. When only the gain lower limit G0 is changed, it is preferable that the curves of λ and μ are changed to include different intervals.
[0102]
In this regard, note that the relationship between gain and λ and μ is different from the function shown in FIG.
[0103]
FIG. 4 shows a multi-channel embodiment of a hearing aid according to the present invention in which each channel generally operates in the same manner as the single-channel embodiment shown in FIG. Corresponding parts in FIGS. 3 and 4 are indicated with the same reference numerals, except that the reference numerals in FIG. 3 are subscripted. For simplicity, only three channels are shown in FIG. However, it should be noted that this hearing aid can include an appropriate number of channels as shown in the figure. For simplicity, the control lines are omitted in FIG.
[0104]
The multi-channel embodiment of the present invention according to FIG. 4 has the same components as the single-channel embodiment shown in FIG. Are output to a first set of adaptive filters and a set of second adaptive filters 11a, 11i, 11n. Each second adaptive filter 11a, 11i, 11n set is filtered into a respective coupling node 12a, 12i, 12n combined with a respective signal 86a, 86i, 86n from the coupling node 9a, 9i, 9n. Supply signal.
[0105]
In the multi-channel embodiment shown in FIG. 4, a more detailed estimate of the transfer function of the feedback path is achieved. Further, signal processing can be performed at a lower sampling frequency in a lower frequency band, a technique known as decimation. Decimation is particularly simple in these filters because the output signal from the second set of adaptive filters is not fed into the main signal path, so no anti-aliasing filter is required for the system. Can be used.
[0106]
The embodiment shown in FIG. 4 can be controlled in the same way as the embodiment shown in FIG. However, in the embodiment shown in FIG. 4, selectively reducing the gain in each individual channel and selecting the adaptation speed of each individual adaptive filter of the first set of adaptive filters 10a, 10i, 10n. It is possible to adjust automatically. For this reason, the gain can be maintained at a high value at a frequency at which feedback resonance is unlikely to occur, and still another advantage is obtained in that the distortion can be maintained at a low level.
[0107]
FIG. 5 shows an embodiment of a multi-channel having the same configuration as that of the embodiment shown in FIG. 4 and operating in the same manner. However, the embodiment shown in FIG. 5 has a first set of adaptive filters constituted by a single adaptive filter 10 and the join node 9 is a single join node, so that It is simple.
[0108]
Many other embodiments are obtained by varying the number of adaptive filters in the channel and first and second adaptive filter sets in the processing unit. Further, the number of channels in the processing device may be different from the number of filters in the first adaptive filter set, and similarly, the number of filters is different from the number of filters in the second adaptive filter set. May be.
[0109]
In particular, it is possible to use a digital signal processing device 7 having a relatively small number of channels and a second set of adaptive filters including more filters. Alternatively, the individual adaptive filters of the first set of adaptive filters may operate on a combination with a channel in the digital signal processor 7, for example two or more of the digital signal processor 7. The channel operates with the same Gmax determined by a particular adaptive filter of the second set of adaptive filters, or one channel of the digital signal processor 7 is driven by an adaptive filter of the second set of adaptive filters You may operate | move using Gmax which is the lowest gain of the two or more gains determined. However, at present, an embodiment with a single first adaptive filter 10 and a set of multi-channel second adaptive filters 11 is preferred.
[0110]
FIG. 11 shows a graph of operating gain as a function of frequency. The upper curve, shown as a solid line, shows the maximum operating gain that can be obtained by the hearing aid according to the present invention without undesired sound generation, while the lower curve, shown as a dashed line, is a known curve The corresponding gain for a hearing aid is shown.
[Brief description of the drawings]
FIG. 1 is a block diagram of a hearing aid according to the present invention.
FIG. 2 is a block diagram of a multi-channel hearing aid, each channel corresponding to the hearing aid shown in FIG.
FIG. 3 is a block diagram of a hearing aid incorporating a measuring device according to the present invention.
4 is a block diagram of a multi-channel hearing aid in which each channel corresponds to the hearing aid shown in FIG.
FIG. 5 is a block diagram of a multi-channel hearing aid with a single band adaptive filter.
FIG. 6 is a block diagram showing an LMS FIR filter that implements an update algorithm according to the present invention.
FIG. 7 is a block diagram illustrating an LMS warp FIR filter that implements an update algorithm in accordance with the present invention.
FIG. 8 is a graph showing a comparison between an impulse response of an FIR filter and an impulse response of a warp type FIR filter.
FIG. 9 is a graph showing a deviation from a desired transfer function of an FIR filter and a warped FIR filter.
FIG. 10 is a diagram illustrating fluctuations in filter coefficients depending on gain in the digital processing device.
FIG. 11 illustrates the maximum possible gain improvement achieved by the present invention.

Claims (25)

Input transducer for converting an acoustic input signal to a first electrical signal,
The first of the electrical signal to bandpass filtering is divided into a first electrical signal set, the first filter bank with bandpass filters,
A first set of coupled nodes is provided that provides a first set of band-filtered electrical signals, and combines the third set of electrical signals with a third set of electrical signals to output a first coupled node output signal;
The first coupling node independently process each set of output signals, and processing apparatus for generating a second electrical signal by adding the processed electrical signals,
An output transducer for converting said second electrical signal into an acoustic output signal,
Have a plurality of bandpass filters for dividing the second electrical signal into a set of bandpass filtering to the second electrical signal, each of the bandpass filter substantially in the plurality of bandpass filter the first filter bank A second filter bank that is identical in nature ,
Estimating the acoustic feedback by generating a set of band filtered second third electrical signals by filtering the electrical signal in accordance with a first set of filter coefficients, with a first filter coefficient A first set of adaptive filters ,
A second adaptive filter with a second filter coefficient that filters the band-filtered second set of electrical signals according to a second set of filter coefficients to generate a fourth set of electrical signals Set of,
The fourth set of electrical signals and the first set of coupled node output signals are combined to generate a fifth set of electrical signals, and the fifth filter is used for adjusting the second filter coefficient. A second set of coupling nodes that inputs the set of electrical signals to the second set of adaptive filters and a first parameter of the acoustic feedback loop of the hearing aid are determined, and a first parameter of the first set of filter coefficients is determined . Equipped with a control device to adjust the adaptive speed of
The control device determines a second parameter of the acoustic feedback loop of the hearing aid, and determines a second filter coefficient having a second adaptive speed higher than the first adaptive speed as the fifth electric signal. Adjust based on the set,
The controller further estimates an acoustic feedback amount based on information from the second set of adaptive filters.
hearing aid. The first of adaptive filter set of at least one adaptive filter, acting on the second electrical signal which is band filtered by a low sampling frequency in the low frequency band, the hearing aid of claim 1. Wherein the first filter bank is configured by a single bandpass filter, hearing aid according to claim 1 or 2. It said second filter bank is configured by a single bandpass filter, the first set of adaptive filter constituted by a single adaptive filter, hearing aid according to claim 1 or 3. Wherein the first set of adaptive filters, the addition to second filtering the electrical signals of, to accommodate the first electric signal, the hearing aid of claim 4. The first coupling node, wherein are those third subtracting the electrical signals of the first electrical signal or, et al., The the reduced signal is supplied to the processing equipment, to claim 5 The hearing aid described. Wherein the first set of adaptive filter is configured to filter the respective second electrical signals which are the band filtering adapted to each of the first electrical signal that is the band filter, according to claim 1 Hearing aids. The first junction node, the third the signals are shall subtracted from each of the first electrical signal that is the band filtering, said the reduced signal is supplied to the processing equipment The hearing aid according to claim 7 . The first parameter is the operation gain of the processing equipment, hearing aid according to any one of claims 1 to 8. The hearing aid according to any one of claims 1 to 8 , wherein the first parameter is a parameter of the first set of adaptive filters. The first parameter is the the signal magnitude at the input of the first of the first adaptive filter of one of the set of adaptive filter is the ratio of the magnitude of the signal at the corresponding output, The hearing aid according to claim 10 . The second parameter is the gain of the processing equipment, hearing aid according to any one of claims 1 to 11. The hearing aid according to any one of claims 1 to 11 , wherein the second parameter is the first convergence speed of the first filter coefficient. The hearing aid according to claim 8 , wherein the first parameter is an operation gain of the processing device, and the second parameter is the second convergence speed of the second filter coefficient. Leakage least mean square algorithm

Further comprising means for updating the filter coefficients according to
Here, ci (n + 1) is an updated value of the i-th filter coefficient, ci (n) is a current value of the i-th filter coefficient, and ci (0) is an initial value of the i-th filter coefficient. , Ui (n) is the (n−i) th sample of the output signal of the processor, e (n) is the current sample of the second electrical signal (86), and λ is leakage. in it, mu is the convergence, and λ and mu determines the first convergence rate, the hearing aid according to any one of claims 1 to 14. Normalized least mean square

Further comprising means for updating the filter coefficients according to
Where u (n) is an N-order vector including the latest N samples of the signal u, c (n) is a vector including the current values of the N filter coefficients, and c (0) is N A vector including initial values of the filter coefficients, c (n + 1) is an updated value of N filter coefficients, and e (n) is a current sample of the second electrical signal (86). The hearing aid according to any one of Items 1 to 14 . Power normalized least mean square algorithm

Further comprising means for updating the filter coefficients according to
The hearing aid according to any one of claims 1 to 14 , wherein α is a predetermined constant that determines the rate at which the Pu estimation changes. Leaky sign least mean square algorithm

Further comprising means for updating the filter coefficients according to
Here, ci (n + 1) is the updated value of the i-th filter coefficient, ci (n) is the current value of the i-th filter coefficient, and ci (0) is the initial value of the i-th filter coefficient. , Ui (n) is the (n−i) th sample of the output signal of the processing unit, e (n) is the current sample of the second electrical signal (86), λ is leakage, μ is the convergence, .mu.s is the product between the code and μ of e (n) signal, lambda and μ determines the first convergence rate, according to any one of claims 1 to 14 hearing aid. Leaky code-sign least mean square algorithm

Further comprising means for updating the filter coefficients according to
Here, ci (n + 1) is the updated value of the i-th filter coefficient, ci (n) is the current value of the i-th filter coefficient, and ci (0) is the initial value of the i-th filter coefficient. , Ui (n) is the (n−i) th sample of the output signal of the processing unit, e (n) is the current sample of the second electrical signal (86), λ is a leak, μ is the convergence coefficient, sgn (ui (n)) is the sign of ui (n), lambda and μ determines the first convergence rate, to an item any of claims 1 to 14 The hearing aid described. It said set of at least one of the first and second adaptive filter comprises a finite impulse response filter, hearing aid according to any one of claims 1 19. It said set of at least one of the first and second adaptive filters comprises a warped shape finite impulse response filter, hearing aid according to any one of claims 1 19. The hearing aid according to any one of claims 1 to 21 , wherein the control device adjusts a second parameter of the hearing aid according to the first parameter and an actual acoustic environment. The hearing aid according to claim 1, wherein at least one of the first and second sets of adaptive filters is a warp type adaptive filter. 24. A hearing aid according to claim 23 , wherein the warp adaptive filter is a warp FIR filter. In a method for suppressing acoustic feedback in a hearing aid,
Input transducer converts the acoustic input signal to a first electrical signal,
The first filter bank is divided into a first set of electrical signals by bandpass filtering the first electrical signal with a bandpass filter,
A first combination node set combines the first electric signal set and the third electric signal set to generate a first combined signal;
Processing apparatus generates a second electrical signal by adding the electrical signals processed by processing separately each of the first combined signal,
Output transducer converts said second electrical signal into an acoustic output signal,
The first of the second filter bank having a plurality of band-pass filters are each substantially identical to the bandpass filter of the filter bank, the second with the second electrical signal to the bandpass filter-ring Divided into sets of electrical signals,
A first set of adaptive filters with first filter coefficients, generates a first set of second third electrical signals by filtering the electrical signal that is the band filter according to a set of filter coefficients estimating acoustic feedback by,
A second set of adaptive filters having a second filter coefficient filters the band-filtered second set of electrical signals according to the set of second filter coefficients to form a fourth set of electrical signals. Generates
A second combination node set combines the fourth electric signal set and the first combination node output signal set to generate a fifth electric signal set, and the second filter coefficient The fifth set of electrical signals is input to the second set of adaptive filters for the adjustment of
The control unit
Determining a first parameter of the acoustic feedback loop of the hearing aid, adjusting a first adaptive speed of the first set of filter coefficients, determining a second parameter of the acoustic feedback loop of the hearing aid; And adjusting a second filter coefficient having a second adaptive speed higher than the adaptive speed based on the information from the second set of adaptive filters based on information from the second set of adaptive filters. Estimate the quantity,
A method of suppressing acoustic feedback that substantially prevents the generation of undesirable sounds.

Images