1380704 九、發明說明 【發明所屬之技術領域】 本發明，係有關於：使用被近接配置之麥克風陣列， 來推定音源之方向，並依據其結果，而將複數方向之音源 同時作收音，並且，可將所收音之聲音，在聲道數或是再 生機器相異之任意的再生系統中作再生的收音方法以及裝 置。 【先前技術】 作爲在音場內之收音裝置，係週知有使用了複數之麥 克風的麥克風陣列裝置。在此麥克風陣列裝置中，爲了削 減麥克風之量的目的，而係提案有：代替實際設置麥克風 並作收音，而以從實際上配置之麥克風所收音的聲音訊號 爲依據，來假定出應爲在假定位置處所被收音的聲音訊號 之技術。專利文獻1之發明，係爲此種技術之代表性者， φ 而係爲在每1維度中藉由2個的麥克風數來對維度方向之 任意位置的收音訊號作推定者。 在此專利文獻1之發明中，如圖6所示一般，係將麥 克風10a、b在軸方向配置2個，並將藉由此所收音之聲 音訊號輸入至受音訊號推定處理部11中。受音訊號推定 處理部11，係將從音源而到達前述2個的麥克風處之音波 近似爲平面波，而將在與麥克風10a、b爲同軸之某位置 處的推定受音訊號，藉由波動方程式來作近似表現，並將 到達前述2個的麥克風之各個處的音波之平均功率假定爲 -5- 1380704 相等’而推定前述波動方程式之依存於音波之到達方向的 係數bcos0，再根據從前述2個的麥克風而來之受音訊號 ’來推定出與該些之麥克風爲同軸上之任意位置的受音訊 號。 [專利文獻1]日本特開2001-45590號公報 【發明內容】 [發明所欲解決之課題] 但是’在視訊會議系統、機器人聽覺等之說話者的方 向推定係被重視的領域中，爲了提昇方向推定之精確度， 麥克風元件之數量係需要增多，又，亦需要有某種程度的 間隔。一般所被檢討之麥克風陣列，包含上述專利文獻1 中之將間隔設爲約3cm的情況，幾乎均係爲將間隔設爲 1 0 0mm以上者。 又’當2聲道雙耳（Binaural )系統的情況時，在頻 率分析或是用以與其作比對之資料庫的利用上等，在藉由 電腦來實現時之演算量、記憶體資源之消耗量或是演算量 上’係有實現上的困難。 進而，對於麥克風陣列之配置，雙耳系統係爲非常容 易受到安裝麥克風之筐體的音響特性影響者。因此，在方 向推定部的安裝作成上，係需要極多的程序。 又，在會議系統中，係在每一說話者處配置麥克風， 並因應於狀況而對聲道作切換，但是，系統之控制主要係 爲手動，又，由於係成爲需要與說話者數量相同之量的麥 -6- 1380704 克風以及傳送路徑（聲道）等，因此系統之規模或成本， 係不得不成爲大規模。 本發明，係爲了解決上述一般之先前技術的問題點而 提案者，其目的，係在於提供一種：使用被近接配置之複 數的麥克風，來對空間上之單數或是複數存在的音源之位 置以及方向作推定，並對於音源所存在之任意的方向，附 加指向性並作收音，藉由此，而能夠以對音源之音響資訊 作強調的形態來作收音的收音方法以及裝置。 [用以解決課題之手段] 本發明，係爲一種將複數之收音用裝置作近接配置， 而在各收音用裝置處，係被連接有因應於再生聲道數之數 量的控制濾波器，並以電腦來實行將從各聲道之控制濾波 器而來之輸出訊號在各聲道中分別作加算並記錄之數位訊 號處理的收音方法，其特徵爲：前述控制濾波器，係在被 近接配置之複數的收音用裝置之周圍音場內部，設定複數 之控制點，並根據實測値，來計算出在此些之控制點與各 收音用裝置間之期望響應函數行列與傳達函數行列，當指 定了前述收音用裝置之指向性的情況時，根據對應於所指 定之指向性的控制點與各收音用裝置間之期望響應函數行 列與傳達函數行列，而決定前述控制濾波器之値，前述數 位訊號處理’係實行：指向性控制處理，係爲了對此控制 濾波器作控制並決定收音時之指向性，而輸入指向性控制 資料：和方向推定處理，係將於每一角度中準備有複數之 1380704 作出任意的指向特性之控制濾波器的濾波係數組， 前述複數之收音用裝置而從前述音場內所收錄之聲 積（Convolutional)，並藉由此來計算出前述音場 向別的音壓分佈，而推定出在前述音場內之音源的 再將此方向推定之結果，輸入至前述指向性控制手 ，此發明，係亦可視爲方法之發明。 在以上一般之形態中，利用藉由方向推定部處 到之方向推定資訊，係成爲能夠自動地對存在音源 而進行附加有指向性之集音。又，例如，就算是在 處之再生，亦可實現對使用本方向推定結果而進行 之音場的再構成、對雙耳音源之變換。 在較理想之形態中，係具備有以下特徵：前述 波器，當將控制濾波矩陣設爲Η(ω)、將期望響 矩陣設爲Α(ω)、將傳達函數設爲C(co)的情況 藉由 Η(ω) = [0(ω) Τ.(3(ω) ]-1(2(ω) T. )來作表現，並藉由對與傳達函數矩陣C((B)之 [〇(ω) T· C(o) ]-lC(w) T 求解而得出。 於此形態中，由於係將對於預先所設定之控制 實側或者是實測値爲依據來演算並求取出其之期望 傳達函數，並以根據此實測値之資料作爲基礎來決 濾波器，因此，不論是對收音用裝置之任何的方向 指向性的情況時，均能夠藉由將構成控制濾波器Η 函數矩陣C的逆矩陣[(：(ω) τ· C(6〇 fCCo) 最小平方法等之近似計算法來求解，而得到近似於 與經由 音作卷 內之方 方向， 段。又 理所得 之方向 遠距離 了集音 控制濾 應函數 時，係 A ( ω 逆矩陣 點而以 響應與 定控制 而賦予 之傳達 )τ以 期望響 -8 - 1380704 應之輸出。 在較理想之形態中，係具備有以下特 定處理’作爲方向推定演算法，係使用任 Forgetting Factor)，而對在前述音場內 之能量作觀測，並藉由預先所設定之音源 來推定音源之存在方向者。 在此形態中，藉由使用方向推定演算 向之音壓分佈，並經由任意之遺忘因 Factor)，而對各方向之能量作觀測，並 源判定的臨限値，來推定音源之存在方向 行自動地對音源之存在方向附加指向性的: 在更爲理想之形態中，係具備有以下 推定處理，係藉由前述方向推定演算法， 係數組之切換並使指向性束旋轉，而推定 的音源之存在方向，再對於所檢測出之音 濾波係數組之選擇。 於此形態中，藉由進行濾波係數組之 束旋轉，而推定出音源之存在方向，再對 源方向，來進行濾波係數組之選擇，而成 地對音源之存在方向附加指向性的集音。 [發明之效果] 若藉由本發明，則能夠提供一種：使 複數的麥克風，來對空間上之單數或是複 徵：前述方向推 意之遺忘因子（ 之各方向的音源 判定的臨限値， 法而計算出各方 子（Forgetting 藉由所設定之音 ，而成爲能夠進 集音。 特徵：前述方向 而進行前述濾波 出在前述音場內 源方向，來進行 切換並使指向性 於所檢測出之音 爲能夠進行自動 用被近接配置之 數存在的音源之 -9- 1380704 位置以及方向作推定，並對於音源所存在之任意的方向， 附加指向性並作收音，藉由此，而能夠以對音源之音響資 訊作強調的形態來作收音的收音方法以及裝置。 【實施方式】 接下來，依據圖面，針對本發明之收音系統的其中一 種實施形態作具體說明。另外，本申請人，係在本發明之 前，已先提案有相關於使用近接配置之麥克風陣列而將指 向性朝向任意之方向來作收音的技術之先前申請（日本特 願2005-351359號）。本發明，係以在此先前申請之發明 中加入有「方向推定部」之點具備有特徵。 Π 實施形態之槪略構成] (1)收音用裝置之其中一例 圖1，係爲展示構成本實施形態中之收音用裝置1之 4個的麥克風Ml〜M4之其中一例者，此些之麥克風Ml 〜M4，係將其收音面朝向相同方向地而被收容在支持器 1 2內。 各麥克風Ml〜M4之間隔，從空間取樣的觀點來看， 係以成爲較所欲收音之音波的4分之1波長爲更短之間隔 爲理想，當將收音之音波設爲聲頻（audio )帶域時，係 以l〇mm左右的間隔來作配置。但是，此尺寸，係並非爲 被限定於本實施形態者，依存於應用之領域，可以設爲從 100mm左右起而至50〜1mm亦無妨。又，進行收音之聲 -10- 1380704 道數（麥克風數），係只要爲2以上即可。 (2 )再生等化電路 經由於圖2所展示之再生等化電路，來對在本發明之 收音系統中所使用的演算法之其中一例作說明。前述各麥 克風Ml〜M4之輸出側，係分別被連接於如圖2所示一般 之再生等化電路。此再生等化電路’係由輸出目標訊號之 期望響應A、和被與此期望響應A作並聯連接之傳達系C 以及控制濾波器Η、和將從前述期望響應A與控制濾波器 Η而來之輸出作加算並輸出誤差e的加算器Σ所構成。 前述期望響應A，係經由以下述之數1式所表現的傳 達函數矩陣Α( ω)而被求取。 [數式1] A ( ω ) = [Αι ( ω ) Α2 ( ω )…Aq ( ω )] 於此，期望響應之矩陣Α(ω)，係如圖3所示一般 ，經由在將麥克風Ml〜Μ4配置於音場空間之收音位置處 的狀態下，於其周圍設定q個的控制點，並實測出從各控 制點而來之脈衝響應而取得之。於此情況，在圖3中，雖 係將麥克風Ml〜M4之周圍360°每隔15°地來作實測， 但是，控制點數係並不限定於此。又，雖係將麥克風Μ1 〜Μ4與各控制點之距離設爲lm，但是，對於此距離，係 亦並不作特別限定。進而，對於此些之作了實測的各控制 -11 - 1380704 點以外之場所處之期望響應，係藉由以內插法等來# f十胃 而取得之。 前述傳達系C’係經由以下述之數2式所表現的傳達 函數矩陣C(co)而被求取。 [數式2] C» ·.· c»- • · * • · · • · « ^Nl (^) * * · ^ΝΜ (ω)_ 於此’ Cn ( ω ) ......... Cim ( ω ) ’係代表第1個的 控制點與各麥克風間之傳達係數，Μ係代表控制點數。又 ’ CN1 ( ω ) .........Cnm( ω )，係代表第Ν個的控制點 與各麥克風間之傳達係數。此傳達函數（：η(ω)......... C i μ ( ω )，係經由對各麥克風Μ 1〜與各控制點間之傳達 特性（衰減或延遲等）作實測而求取之。 前述控制濾波器Η，係根據前述期望響應傳達函數a (ω)與傳達函數矩陣C(co)，而藉由下述之數3式來 求取出。 [數式3] Η(ω) = [0(ω) τ· 0(ω) ]''〇(〇)) τ· A ( ω ) 亦即是，如同由前述之各式而可明顯得知一般，在圖 -12- 1380704 2之再生等化電路中，由於係經由加算器ς而從期望響應 傳達函數Α(ω)來將被包含在控制濾波器η中的Α(ω )作減算，因此，爲了得到使從再生等化電路所輸出之誤 差e成爲最小的控制濾波器η，係成爲只要將構成控制濾 波器Η之傳達函數矩陣c的逆矩陣[C((a) T· C((〇) ]-*C (ω ) T以最小平方法等之近似計算法來求解即可。於此 情況，基於最小平方法之解法，係可適用最陡降下法（ Steepest descent method)等之各種數値計算法》 [2.實施形態之具體構成] (1 )全體構成 本實施形態之收音系統，係如圖4所示一般，藉由對 於如前述一般之複數的麥克風以及被連接於複數之麥克風 之輸出側的控制濾波器Η，而將監聽系統以及再生系統作 組合來構成。另外，在圖1中，作爲收音裝置，雖係展示 了 4個的麥克風，但是，在圖4之實施形態中，係將收音 用之麥克風數設爲Μ，將再生聲道數設爲Ν。 於圖4中，1係爲收音用裝置，2係爲數位訊號處理 部，3係爲監聽處理部，4係爲再生處理部，收音用裝置1 ，係具備有收音用之麥克風幻〜“。 (2)數位訊號處理部之構成 數位訊號處理部2，係具備有被連接於各收音用麥克 風11〜Ιμ之輸出側的控制濾波器Η! !〜ΗΜΝ。亦即是，在 -13- 1380704 各收音用麥克風Ii〜IM處，係分別被連接有對應於再生聲 道數N之控制濾波器H。又，被連接於各麥克風之各聲道 用的控制濾波器Η，係被連接於各再生聲道用之加算器 Σ 1 〜Σ ν。 在數位訊號處理部2中之各控制濾波器Hu〜ΗΜΝ處 ，係被連接有用來輸入用以決定收音用麥克風Ι-Ιμ之指 向性的控制資料之指向性控制部2 1。亦即是，此指向性控 制部21，係爲了在經由各收音用麥克風h〜IM而從音場 內所收錄之聲音中，對從所期望之方向與位置所發出之聲 音作強調而進行收音，而對於數位訊號處理部2，將該方 向與位置作爲控制資料而輸入。 (2-1 )指向控制部之構成 在此指向控制部21處，係由使用者來藉由編碼器或 是鍵盤等而將控制資料直接作輸入，或是經由電腦程式來 輸入經時變化之控制資料。於此情況，在所輸入之指向性 的控制資料中，係將對經由後述之方向推定部22的處理 所得到之期望響應作了測定的前述q個的控制點之1個場 所或是複數場所作指定。 例如，當輸出聲道係爲1聲道的情況時’只要指定1 個場所的控制點即可，而當多聲道的情況時，則係將對應 於輸出聲道數與方向之數量與方向的控制點，作爲控制資 料而輸入。圖5，係爲展示：作爲5聲道再生系統，而對 於圖1所示之麥克風Ml〜M4周圍的5方向來使其持有麥 -14- 1380704 克風之指向性，並對該方向之聲音作強調而進行收音的狀 態者。 此指向性控制部21，若是經由後述之方向推定部22 的處理而被輸入有控制點，則係根據由實測値所得到之相 關於該控制點的期望響應傳達函數矩陣Α(ω)與傳達函 數矩陣〇(ω)，並依循前述（2)所示之演算法，而進 行對各控制濾波器Hu〜HMN之値作決定的演算，並將其 演算結果輸出至數位訊號處理部2。 (2-2)方向推定部之構成 接下來，針對身爲本發明之特徵構成的方向推定部22 之構成作說明。 方向推定部22，係具備有：濾波係數組，其係在指向 性控制部2 1的處理之前，預先將做出任意之指向特性的 濾波器Η對每一角度中而準備了複數個。 方向推定部22，係藉由將此濾波係數組與經由各收音 用麥克風Ιι〜Ιμ而從音場內所收錄的聲音作卷積，而計算 出各方向之音壓分佈，並進行音源之方向推定者。又，係 爲藉由對濾波器Η作高速切換而恆常對驅動之控制濾波器 作限定，來在抑制演算量之同時亦進行所有方向之音源音 壓分佈的算出者。而後’將此方向推定之結果，對於數位 訊號處理部2而作爲控制資料來輸入。以下，作爲其具體 構成，針對（Α)濾波係數組、（Β)方向推定之手法、（ C)推定演算法，作具體之說明。 -15- 1380704 (A )關於濾波係數組 所謂濾波係數，係爲用以在各方向而形成指向性的係 數，對於任意之一方向，係準備有一組。此組，係將預先 被設計者作儲存。理論上，對於全方向，係以準備無限個 的係數組爲理想，但是，實際上，由於會有記億體等之硬 體資源的各種之限制，因此，係以空出一定間隔的狀態， 來準備對應於複數方向之複數的濾波係數組。 在本實施形態中，係如圖6之示意圖所示一般，設定 爲：對於特定之8個方向，而準備8個的濾波係數組，並 將其與從麥克風陣列而來之輸入作卷積。 (B)方向推定之手法 藉由對上述一般之濾波係數組作切換，能夠朝向任意 之方向而形成指向性。利用此點，方向推定部22，係逐漸 地對濾波係數組作切換，並計測出在任意之方向處的音場 之能量。亦即是，藉由使方向推定部之指向性作即時（ realtime )的變化，而得到在單位時間內之所有方向的音 壓分佈。 此，係恆常將驅動之控制濾波器Η ! !〜HMN限定於少 數，而在抑制演算量之同時亦進行所有方向之音壓分佈的 觀測者。 藉由上述一般之濾波係數組與方向推定之手法，本實 施形態之方向推定部22，係對在預先所設定之複數的方向 -16- 1380704 處之音場的能量作觀測，並藉由以下之推定演算法.’而進 行音源之方向推定。 (c)方向推定演算法 方向推定部22之方向推定演算法，係如同在圖7之 區塊圖中所示一般，使用任意之遺忘因子（Forgetting Factor )，而對各方向之能量作觀測，並藉由所設定之音 源判定的臨限値，來推定音源之存在方向者。遺忘因子之 値以及音源判定之臨限値，係以因應於使用用途來作最適 當之設定爲理想。 藉由此方向推定演算法，如同圖8之示意圖所示一般 ，進行濾波係數組之切換並使指向性束旋轉，而推定出音 源之存在方向，再對於所檢測出之音源方向，來進行濾波 係數組之選擇。 (3)其他構成 監聽處理部3,係具備有如同耳機或是2聲道揚聲器 —般之2聲道的監聽用輸出部0^02。在此監聽用輸出 部Oi'02處，從前述各再生聲道之加算器Σ,〜ΣΝ而來 的訊號’係經由假想音源再生處理部31而被輸出。。 亦即是’假想音源再生處理部31，係將從各再生聲道 之加算器Σ!〜Σν而來的訊號，分割爲左右之揚聲器或者 是耳機用，並使此被分割之左右的訊號，分別通過控制濾 波器Si' C1〜Sn、cn，而後，將各再生聲道之右側的控制 -17- 1380704 瀘波器S1〜Sn的輸出，經由加算器Σ(π而作加算並輸出至 監聽用輸出部〇1，又，將各再生聲道之左側的控制濾波 器1〜（：„的輸出，經由加算器Σ〇2而作加算並輸出至監 聽用輸出部〇2。 於此情況，前述控制濾波器S,、c,〜sn、cn，係爲因 應於作爲監聽用輸出部0^02而使用之揚聲器或耳機等 的裝置而具備有相異之濾波係數者，藉由此，而在各裝置 中產生適合於聽取者之兩耳所收聽的訊號0 又，在監聽處理部3中，當經由前述數位訊號處理部 2而指定了所應收音之特定控制點的情況時，係被設置有 用以指定應對何者之聲道的聲音作監聽一事的聲道指定部 32。此聲道指定部32，係爲從由數位訊號處理部2所輸出 之各聲道的訊號中，僅對進行監聽之聲道的訊號作指定， 並將其輸入至假想音源再生處理部31處者。 再生處理部4，係具備有將從前述數位訊號處理部2 處之各聲道用的加算器Σ,〜ΣΝ而來的訊號作輸出之各聲 道的再生輸出部Α-Ον。此再生輸出部Oi-ON，係進而 被連接於立體聲系統、5_1聲道環繞聲系統、假想音源再 生處理部等之任意的再生系統之輸入。 [3.實施形態之作用] (1 )指向性控制部之控制點的設定 具備有以上一般之構成的本實施形態之收音系統的作 用，係如下述所示一般。首先，在收音之前，將複數之各 -18- 1380704 收音用裝置以近接之狀態而配置在音場空間內，並於其周 圍設定複數之控制點。於該狀態下，藉由將從各控制點所 發出之聲音以各收音裝置來作收錄，而從測定値來求取出 各控制點與各收音用裝置間之期望響應函數矩陣Α(ω) 與傳達函數矩陣C(w)，並將此些儲存在指向性控制部 2 1內。 另一方面，在進行再生處理時，係對進行幾聲道之再 生一事作決定，並在再生處理部4中準備聲道數量之再生 裝置，而將此些之再生裝置連接於被設置在數位訊號處理 部2中之各聲道的再生輸出部Ch〜0N。又，控制濾波器 Α'Ημ，亦在每一近接配置之各收音裝置卜〜“中，準 備再生聲道數量的份量。 另外，再生聲道數，係並不需要事先作決定，而亦可 將經由各收音裝置所收錄之聲音儲存在記憶裝置中，並在 決定了再生聲道述之後，再準備具備有所需要之數量的控 制濾波器與加算器的數位訊號處理部2、和再生用裝置。 在此種狀態下，經由各收音用裝置h〜IM所收錄的聲 音，係被輸入至方向推定部22處。 (2)方向推定部之處理 於此，方向推定部22，係藉由將於每一角度中準備有 複數之作出任意之指向特性的濾波器Η之濾波係數組，與 經由各收音用麥克風h〜Ιμ而從音場內所收錄的聲音作卷 積，來進行音源之方向推定。使用圖9之流程圖，來對此 -19- 1380704 處理作說明。 如圖9所示一般，方向推定部22，首先，係藉由對預 先所準備之濾波係數作切換，而進行指向性束之旋轉（參 考圖6之示意圖），並計測出在任意之方向上的音場之能 量(STEP1 )。 接下來，根據在STEP1中之計測，而使用方向推定演 算法來檢測出音源之方向（STEP2 )。具體而言，方向推 定部22，係使用任意之遺忘因子（Forgetting Factor)， 而對各方向之能量作觀測，並藉由所設定之音源判定的臨 限値，來推定音源之存在方向》接下來，對是否檢測出有 音源之存在方向一事作確認（STEP3 )，當並未檢測出存 在有音源之方向的情況時（NO )，則回到STEP2，並反覆 進行STEP2〜STEP 3之處理。另一方面，當檢測出了存在 有音源之方向時（YES )，則前進至STEP4。 若是檢測出了音源之方向（STEP3之YES )，則選擇 出對該方向作強調的濾波係數組（STEP4 )，並將此輸入 至指向性控制部21中（STEP5 )。 (3 )指向性控制部之處理 若是藉由如上述一般之方向推定部22的處理，而對 指向性控制部21輸入了應對何一方向之聲音作強調作收 音一事，則在指向性控制部21中，係根據所輸入之方向 與位置（與收音用裝置間之距離），而選擇預先實測有期 望響應函數以及傳達函數（或者是從實測値來作演算並求 -20- 1380704 取）之控制點，並呼叫出該控制點q之期望響應函數矩陣 以及傳達函數矩陣，再藉由將此些代入至前述數3式中’ 來對控制濾波器Hu〜HMN之値作演算並求取之。 此時，由於各收音用裝置Ιι〜Im與控制點q間之距離 或方向係爲相異，因此，其之期望響應函數以及傳達函數 亦係分別相異，又，當再生聲道係爲複數的情況時，由於 在各聲道中之給予收音用裝置的指向性的方向（收音用裝 置所強調並收音之方向）亦係爲相異，因此，各控制濾波 器之値亦成爲相異 如此這般，若是決定了各控制濾波器之値，則經由此 些之控制濾波器Hh〜HMN，在每一各聲道中，於各收音 用裝置之聲音中僅有所期望之方向的聲音被作強調。而後 ’從各控制濾波器而來之訊號，係在每一各聲道中，經由 加算器Σ!〜ΣΝ而被作加算，而該些係從各聲道之輸出部 再生輸出部0,〜0Ν而被輸出至各聲道之再生用裝置。 接下來，在本實施形態中，爲了進行再生聲道之監聽 ’係對於監聽處理部3，而從聲道指定部32來指定欲進行 監聽之聲道。如此一來，從由被設置在數位訊號處理部中 之各聲道的加算器而來之訊號中，僅選擇所期望 之聲道的訊號，而該訊號，係經由控制濾波器Sl、Cl〜Sn ' Cn ’而被輸出至身爲監聽用之再生裝置的2聲道之揚聲 器或是耳機處。此時，因應於作輸出之再生裝置，來對前 述控制濾波器S!，（^〜3η，Cn之係數作設定，藉由此， 不論再生用裝置之種類爲何，均能夠得到最適當之輸出。 -21 - 1380704 [4.實施形態之效果] 在上述一般之本實施形態中，方向推定部22,係藉由 將身爲用以在各方向形成指向性之係數的濾波係數組與經 由各收音用麥克風h〜IM而從音場內所收錄的聲音作卷積 ，並使用方向推定演算法而計算出各方向之音壓分佈，而 經由任意之遺忘因子，來在各方向對能量作觀測，並藉由 所設定之音源判定的臨限値，而能夠進行音源之存在方向 的推定》 而後，利用藉由此方向推定部22所得到之方向推定 資訊，係成爲能夠自動地對音源之存在方向而進行附加有 指向性之集音。又，就算是在遠距離處之再生，亦可實現 對使用本方向推定結果而進行了集音之音場的再構成、對 雙耳音源之變換。 經由使用此種方向推定部，例如，係成爲可利用於機 器人之音源方向推定，亦即是，成爲可利用於：利用在機 器人耳中之音源方向推定感測器、雞尾酒會效果（ Cocktail party, effect)之實現等。又，可適用於多聲道錄 音系統，亦即是，可適用於5.1 ch錄音等。進而，可利用 於高臨場感通訊系統，例如，可利用於機器人或會議系統 等之中，就算是自己並不前往該場所，亦可在.遠距離處創 造出與該場所相同之音場空間。 又，由於係對於預先所設定之控制點q而將該期望響 應與傳達函數作實測或者是以實測値爲依據來演算並求取 -22- 1380704 出’並以根據此實測値之資料作爲基礎來決定控制濾波器 ，因此，不論是對收音用裝置之任何的方向而賦予指向性 的情況時，均能夠藉由將構成控制濾波器Η之傳達函數矩 陣C的逆矩陣[(：(ω)τ·(：(ω) ]·丨C ( ω ) τ以最小平 方法等之近似計算法來求解，而得到近似於期望響應之輸 出。 又’在本實施形態中’由於係構成爲將從數位訊號處 理部2而來之輸出導引至監聽處理部3，並輸入輸出至2 聲道之再生用裝置，因此’不論是對何者之再生聲道的輸 出’均只要對被設置在監聽處理部3處之聲道指定部32 作操作’即可與其他之聲道音作明確的區別並聽取。當然 ，於此情況中，雖然亦能夠僅對單一之再生聲道的聲音作 監聽，但是，亦可將從加算器Σ!〜ΣΝ所輸出之複數的聲 道之聲音，同時地輸出至監聽用裝置處。 【圖式簡單說明】 [圖1]展示在本發明中所使用之麥克風的構成例之圖 ，（A )係爲側面圖，（B )係爲正面圖。 [圖2]展示用以得到構成本發明之收音系統的控制濾 波器Η之演算法的再生等化電路圖。 [圖3]展示在本發明中將期望響應在音場空間內作了 設定的狀態之圖。 [圖4]展示本發明之收音系統的其中一種實施形態之 區塊圖。 -23- 1380704 [圖5]展示在麥克風之周圍的5個方向設定了指向性 的狀態之圖。 [圖6]展示在本發明中形成了濾波係數組所致之指向 性的狀態之圖。 [圖7]展示在本發明中之用以產生任意的遺忘因子之 其中一種實施形態之區塊圖。 [圖8]展示在本發明中之對濾波係數組作切換所致之 指向性數之旋轉狀態之圖。 [圖9]展示在本發明中之方向推定部的處理之流程圖 〇 [圖10]展示先前之收音系統的其中一例之區塊圖。 【主要元件符號說明】1380704 IX. Description of the Invention [Technical Field] The present invention relates to estimating a direction of a sound source using a microphone array that is closely arranged, and, according to the result, simultaneously collecting sound sources in a plurality of directions, and The sound collection method and apparatus for reproducing the sound of the sound in any of the reproduction systems in which the number of channels or the reproduction machine is different. [Prior Art] As a sound pickup device in a sound field, a microphone array device using a plurality of microphones is known. In the microphone array device, in order to reduce the amount of the microphone, it is proposed to replace the actual setting of the microphone and to receive the sound, and to assume that it should be based on the sound signal received from the actually arranged microphone. The technique of assuming an audio signal at a location. The invention of Patent Document 1 is a representative of such a technique, and φ is an estimate of the sound pickup signal at any position in the dimension direction by the number of microphones in each dimension. In the invention of Patent Document 1, as shown in Fig. 6, in general, two microphones 10a and 10b are arranged in the axial direction, and the sound signals received therefrom are input to the sound receiving signal estimation processing unit 11. The sound signal estimation processing unit 11 approximates a sound wave from the sound source to the two microphones to a plane wave, and estimates a sound signal at a position coaxial with the microphones 10a and b, by the wave equation. For the approximate performance, the average power of the sound waves reaching each of the two microphones is assumed to be -5 - 1380704 equal ', and the coefficient bcos0 of the wave equation depending on the direction of arrival of the sound wave is estimated, and then according to the above 2 The microphones are derived from the sound signal 'to estimate the sound signal at any position coaxial with the microphones. [Patent Document 1] JP-A-2001-45590 [Summary of the Invention] [Problems to be Solved by the Invention] However, in the field where the direction of the speaker of the video conferencing system and the robot auditory is emphasized, in order to improve The accuracy of the direction estimation, the number of microphone components is required to increase, and a certain degree of separation is required. In general, the microphone array to be examined includes the case where the interval is set to about 3 cm in the above Patent Document 1, and the interval is set to be more than 100 mm. In the case of the 2-channel binaural system, the amount of calculation and memory resources are realized by computer when the frequency analysis or the use of the database for comparison is used. The consumption or the amount of calculations is difficult to achieve. Furthermore, for the configuration of the microphone array, the binaural system is very susceptible to the acoustic characteristics of the housing in which the microphone is mounted. Therefore, in the installation of the direction estimating unit, a large number of programs are required. Moreover, in the conference system, the microphone is arranged at each speaker, and the channel is switched according to the situation. However, the control of the system is mainly manual, and since the system needs to be the same as the number of speakers. The amount of wheat -6 - 1380704 grams of wind and transmission path (channel), etc., so the size or cost of the system, has to become a large-scale. The present invention has been made in order to solve the problems of the above-mentioned general prior art, and an object thereof is to provide a position of a sound source in a singular or plural number in space using a plurality of microphones arranged in close proximity and The direction is estimated, and the directivity is added to any direction in which the sound source is present, and the sound collection method and apparatus for collecting the sound information of the sound source can be made. [Means for Solving the Problem] The present invention is a method in which a plurality of sound pickup devices are arranged in close proximity, and a control filter corresponding to the number of reproduced channels is connected to each sound pickup device, and A computer is used to implement a digital signal processing method in which an output signal from a control filter of each channel is added and recorded in each channel, and the control filter is configured in proximity. Within the surrounding sound field of the plurality of sound collecting devices, a plurality of control points are set, and based on the actual measurement, the desired response function row and the transmission function row between the control points and the respective sound collecting devices are calculated, when specified In the case of the directivity of the above-mentioned sound pickup device, the digital signal is determined based on the desired response function between the control point corresponding to the specified directivity and each of the sound pickup devices, and the transmission function row is determined. Signal processing 'system implementation: directional control processing, in order to control this control filter and determine the directivity of the radio, and the input finger Sexual control data: and direction estimation processing, which is to prepare a complex filter coefficient group for each of the angles of 1380704 to make an arbitrary directional characteristic control filter, which is included in the above-mentioned sound field by the above-mentioned plural sound collecting device. Convolutional, by which the sound field distribution to the other sound pressure is calculated, and the result of estimating the direction of the sound source in the sound field is estimated and input to the directional control hand. This invention is also considered to be an invention of the method. In the above-described general form, the information is estimated by the direction in which the direction estimating unit is located, and it is possible to automatically add the directivity to the existing sound source. Further, for example, even if it is reproduced at the same place, the reconstruction of the sound field using the estimation result of the direction and the conversion of the binaural sound source can be realized. In a preferred embodiment, the filter has a characteristic that the control filter matrix is set to Η(ω), the desired response matrix is set to Α(ω), and the transfer function is set to C(co). The situation is expressed by Η(ω) = [0(ω) Τ.(3(ω) ]-1(2(ω) T. ), and by the pair of transitive function matrices C((B)[ 〇(ω) T· C(o) ]-lC(w) T is obtained by solving. In this form, since the system will calculate and extract the pre-set control real side or actual measurement 値The expectation conveys the function, and based on the data of the actual measurement, the filter is determined. Therefore, whether it is any direction directivity of the radio device, the control filter Η function matrix can be constructed. The inverse matrix [C: (ω) τ · C (6〇fCCo) minimum flat method of C is solved by the approximate calculation method, and the direction is approximated by the direction in the volume of the volume. When the long-range sound control filter function is used, the system A ( ω inverse matrix point and the response is given by the control) τ to expect the sound -8 - 1380704 should In the preferred form, the following specific processing is used as the direction estimation algorithm, which uses the Forgetting Factor, and the energy in the sound field is observed and is set by the previously set sound source. Predicting the direction of existence of the sound source. In this form, by using the direction estimation calculus to the sound pressure distribution, and by any forgetting factor (Factor), the energy of each direction is observed, and the threshold of the source is determined, It is estimated that the direction of existence of the sound source automatically adds directivity to the direction of existence of the sound source: In a more ideal form, the following estimation process is performed, and the coefficient group is switched and pointed by the above-described direction estimation algorithm. The beam is rotated, and the direction of the estimated sound source is present, and then the selection of the detected sound filter coefficient group is selected. In this form, the direction of existence of the sound source is estimated by performing the beam rotation of the filter coefficient group, and then In the source direction, the selection of the filter coefficient group is performed, and the directional direct sound is added to the existence direction of the sound source. [Effect of the invention] Ming, can provide a kind of: to make a plurality of microphones, to the singular or recursive in space: the forgotten factor of the above-mentioned direction (the threshold of the sound source in each direction), calculate the parties ( Forgetting is capable of collecting sound by the set sound. Features: The above-mentioned direction is filtered and the source direction of the sound field is switched, and the directivity is made to be automatically used for the detected sound. The position and direction of the -9- 1380704 of the sound source existing in the proximity configuration are estimated, and the directionality is added to any direction in which the sound source exists, and the sound is added, thereby being able to emphasize the sound information of the sound source. The form of the radio method and apparatus for radio reception. [Embodiment] Next, one embodiment of the sound pickup system of the present invention will be specifically described based on the drawings. Further, the present applicant has previously proposed a prior art application (Japanese Patent Application No. 2005-351359) relating to the use of a microphone array of a proximity arrangement to direct the directivity in an arbitrary direction. The present invention is characterized in that the "direction estimating unit" is added to the invention of the prior application.槪 槪 构成 ] ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 此 此 此Ml to M4 are housed in the holder 1 2 with their sound receiving faces facing in the same direction. The interval between the microphones M1 to M4 is ideal from the viewpoint of spatial sampling, and is preferably shorter than the one-fourth wavelength of the sound wave to be sounded. When the sound wave is set to audio (audio) When the band is used, it is configured at intervals of about 10 mm. However, this size is not limited to the embodiment, and may be from about 100 mm to 50 to 1 mm depending on the field of application. In addition, the sound of the radio -10- 1380704 (the number of microphones) can be 2 or more. (2) Regeneration equalization circuit An example of an algorithm used in the radio system of the present invention will be described with respect to the reproduction equalization circuit shown in Fig. 2. The output sides of the respective microphones M1 to M4 are connected to a general regenerative circuit as shown in Fig. 2, respectively. The regenerative equalization circuit 'is derived from the expected response A of the output target signal, and the communication system C and the control filter 被 connected in parallel with the expected response A, and will be derived from the aforementioned desired response A and the control filter. The output is composed of an adder that adds up and outputs an error e. The aforementioned expected response A is obtained via the propagation function matrix Α(ω) expressed by the following equation 1. [Expression 1] A ( ω ) = [Αι ( ω ) Α 2 ( ω )...Aq ( ω )] Here, the matrix 期望(ω) of the desired response is as shown in FIG. 3, generally via the microphone M1. When the Μ4 is placed in the sound receiving position of the sound field space, q control points are set around the sound field, and the impulse responses from the respective control points are actually measured and obtained. In this case, in Fig. 3, although the surroundings of the microphones M1 to M4 are measured at intervals of 360° every 15°, the number of control points is not limited thereto. Further, although the distance between the microphones Μ1 to Μ4 and the respective control points is lm, the distance is not particularly limited. Further, the expected response at each of the places other than the control -11 - 1380704 points which were actually measured is obtained by interpolation or the like. The communication system C' is obtained by the transmission function matrix C(co) expressed by the following equation 2. [Expression 2] C» ··· c»- • · * • · · · · « ^Nl (^) * * · ^ΝΜ (ω)_ This is 'Cn ( ω ) ....... .. Cim ( ω ) ' represents the communication coefficient between the first control point and each microphone, and the system represents the number of control points. Further, ' CN1 ( ω ) ... ... Cnm( ω ) represents the transmission coefficient between the control point of each of the first and each microphone. This transfer function (: η(ω)......... C i μ ( ω ) is measured by the communication characteristics (attenuation or delay, etc.) between each microphone Μ 1 to each control point. The control filter 传达 transmits the function a (ω) and the transfer function matrix C(co) according to the expected response, and is obtained by the following equation 3. [Expression 3] Η ( ω) = [0(ω) τ· 0(ω) ]''〇(〇)) τ· A ( ω ) is, as is evident from the above equations, in general, in Figure -12- In the regenerative circuit of 1380704, the Α(ω) included in the control filter η is subtracted from the expected response transfer function ω(ω) via the adder ,, so that the slave is reproduced. The control filter η at which the error e output from the equalization circuit becomes the smallest is an inverse matrix of the transfer function matrix c constituting the control filter [ [C((a) T· C((〇) )]**C (ω ) T can be solved by an approximate calculation method such as the least square method. In this case, based on the solution of the least squares method, various numbers of 値 calculations such as the steepest descent method can be applied. [2. Specific configuration of the embodiment] (1) The overall configuration of the radio system of the present embodiment is as shown in Fig. 4, and is generally provided by a plurality of microphones as described above and an output side connected to a plurality of microphones. The control filter Η combines the monitoring system and the reproduction system. In addition, in Fig. 1, although four microphones are shown as the sound pickup device, in the embodiment of Fig. 4, The number of microphones for radio reception is set to Μ, and the number of reproduction channels is set to Ν. In Fig. 4, 1 is a radio device, 2 is a digital signal processing unit, 3 is a monitor processing unit, and 4 is a regenerative process. The radio device 1 includes a microphone for sound pickup. (2) The digital signal processing unit 2 of the digital signal processing unit is provided with an output side connected to each of the sound pickup microphones 11 to Ιμ. The control filter Η! !~ΗΜΝ, that is, the control filters H corresponding to the number of reproduced channels N are respectively connected to the respective radio microphones Ii to IM at -1 to 1380704. Each microphone The control filter Η is connected to the adders Σ 1 to Σ ν for each of the reproduced channels. The control filters Hu to ΗΜΝ in the digital signal processing unit 2 are connected for input. The directivity control unit 21 that determines the directivity control data of the microphone Ι-Ιμ, that is, the directivity control unit 21 is included in the sound field for passing through the microphones h to IM for each of the sound pickups. In the sound, the sound emitted from the desired direction and position is emphasized and the sound is received, and the digital signal processing unit 2 inputs the direction and position as control data. (2-1) The configuration of the pointing control unit is directed to the control unit 21, and the user directly inputs the control data by means of an encoder or a keyboard or the like, or inputs the change over time via a computer program. Control data. In this case, in the control data of the directivity to be input, one or more of the q control points for measuring the expected response obtained by the process of the direction estimating unit 22, which will be described later, are measured. Designated. For example, when the output channel is 1 channel, 'as long as the control point of 1 place is specified, and when multi-channel is used, the number and direction of the output channel number and direction will be corresponding. The control point is entered as control data. Fig. 5 is a view showing a directivity of the wheat-14-1380704 gram in the direction of the five directions around the microphones M1 to M4 shown in Fig. 1 as a 5-channel reproducing system, and in the direction of The state in which the sound is emphasized and the sound is collected. When the control point is input via the process of the direction estimating unit 22 to be described later, the directivity control unit 21 transmits the function matrix Α(ω) and the communication based on the expected response related to the control point obtained by the actual measurement. The function matrix 〇(ω) is subjected to the calculation of the control filters Hu to HMN in accordance with the algorithm shown in the above (2), and the calculation result is output to the digital signal processing unit 2. (2-2) Configuration of Direction Estimation Unit Next, the configuration of the direction estimation unit 22 which is a characteristic feature of the present invention will be described. The direction estimating unit 22 is provided with a filter coefficient group which is prepared in advance for each angle by a filter 做出 which has an arbitrary directivity characteristic before the processing by the directivity control unit 21. The direction estimating unit 22 calculates the sound pressure distribution in each direction by convolving the filter coefficient group with the sound collected from the sound field via the respective microphones 收ι to Ιμ, and performs the direction of the sound source. Presumer. Further, the control filter for driving is constantly limited by switching the filter at a high speed, and the calculation of the sound pressure distribution in all directions is performed while suppressing the calculation amount. Then, the result of estimating this direction is input to the digital signal processing unit 2 as control data. Hereinafter, as a specific configuration, a (Α) filter coefficient group, a (Β) direction estimation method, and a (C) estimation algorithm will be specifically described. -15- 1380704 (A) Regarding the filter coefficient group The filter coefficient is a coefficient for forming directivity in each direction, and one set is prepared for any one of the directions. This group will be stored by the designer in advance. In theory, it is desirable to prepare an infinite number of coefficient groups for the omnidirectional direction. However, in reality, since there are various restrictions on the hardware resources such as the billions of bodies, the state is vacant at a certain interval. A filter coefficient group corresponding to the complex number of the complex direction is prepared. In the present embodiment, as shown in the schematic diagram of Fig. 6, generally, eight filter coefficient groups are prepared for the specific eight directions, and are convoluted with the input from the microphone array. (B) Direction estimation method By switching the above-described general filter coefficient group, directivity can be formed in an arbitrary direction. At this point, the direction estimating unit 22 gradually switches the filter coefficient group and measures the energy of the sound field in an arbitrary direction. That is, the sound pressure distribution in all directions per unit time is obtained by making the directivity of the direction estimating unit a real time change. In this case, the constant control drive Η ! ! ~ HMN is limited to a small number, and the observer who performs the sound pressure distribution in all directions while suppressing the calculation amount is also performed. According to the above-described general filter coefficient group and direction estimation method, the direction estimating unit 22 of the present embodiment observes the energy of the sound field at a predetermined direction -16 - 1380704 in the preset direction, and by the following The estimation algorithm. 'And the direction of the sound source is estimated. (c) direction estimation algorithm for estimating the direction of the algorithm direction estimating unit 22, as shown in the block diagram of Fig. 7, using an arbitrary forgetting factor (Forgetting Factor), and observing the energy in each direction, And by the threshold of the set sound source, the direction of existence of the sound source is estimated. The threshold of the forgetting factor and the threshold of the sound source determination are ideal for the most appropriate setting for the intended use. By deriving the algorithm in this direction, as shown in the schematic diagram of FIG. 8, the switching of the filter coefficient group is performed and the directional beam is rotated, and the direction of existence of the sound source is estimated, and then the direction of the detected sound source is filtered. The choice of coefficient group. (3) Other configuration The monitoring processing unit 3 is provided with a monitoring output unit 0^02 having two channels like a headphone or a two-channel speaker. At the monitor output unit Oi'02, the signal ’ from the adder Σ of the respective reproduced channels is output via the virtual sound source reproduction processing unit 31. . In other words, the imaginary sound source reproduction processing unit 31 divides the signal from the adder Σ!~Σν of each reproduction channel into left and right speakers or headphones, and divides the left and right signals. By controlling the filters Si' C1 to Sn and cn respectively, the outputs of the control -17-1380704 choppers S1 to Sn on the right side of each reproduction channel are added to the monitor via the adder Σ (π) In the output unit 〇1, the output of the control filters 1 to (: „ on the left side of each reproduction channel is added to the monitor output unit 〇2 via the adder Σ〇2. In this case, The control filters S, c, sn, and cn are provided with different filter coefficients in response to a device such as a speaker or an earphone used as the monitoring output unit ○02. In each device, a signal 0 suitable for listening to both ears of the listener is generated, and in the case where the monitoring processing unit 3 specifies a specific control point of the sound to be received via the digital signal processing unit 2, Set the sound to specify which channel to respond to a channel specifying unit 32 for monitoring, and the channel specifying unit 32 specifies only the signals of the channels to be monitored from the signals of the channels output by the digital signal processing unit 2, and It is input to the virtual sound source reproduction processing unit 31. The reproduction processing unit 4 is provided with signals for outputting the signals from the adder ΣΝ, ΣΝ from the respective channels of the digital signal processing unit 2 The reproduction output unit Α-Ον of the track is further connected to an input of an arbitrary reproduction system such as a stereo system, a 5_1 channel surround sound system, or a virtual sound source reproduction processing unit. (1) The setting of the control point of the directivity control unit includes the function of the radio system of the present embodiment having the above-described general configuration, and is generally as follows. First, before the radio reception, each of the plural numbers is - 18- 1380704 The radio device is placed in the sound field space in a close-up state, and a plurality of control points are set around it. In this state, the sounds emitted from the control points are made by the respective sound pickup devices. Inclusion, From the measurement, the desired response function matrix Α(ω) and the transfer function matrix C(w) between each control point and each of the sound pickup devices are extracted, and these are stored in the directivity control unit 21. On the other hand, when the reproduction processing is performed, the reproduction of several channels is determined, and the reproduction device of the number of channels is prepared in the reproduction processing unit 4, and the reproduction devices are connected to be set in the digital signal processing. The reproduction output units Ch to 0N of the respective channels in the unit 2. Further, the control filter Α'Ημ is also prepared for each of the sound collection devices in the proximity arrangement, and the number of reproduction channels is prepared. The number of channels does not need to be determined in advance, but the sounds recorded by the respective radios can be stored in the memory device, and after the reproduction of the channel is determined, the required number of controls is prepared. The digital signal processing unit 2 of the filter and the adder and the reproducing device. In this state, the sounds recorded by the respective sound collecting devices h to IM are input to the direction estimating unit 22. (2) Process of the direction estimating unit The direction estimating unit 22 is configured to have a filter coefficient group of a filter 作出 having an arbitrary directivity characteristic at each angle, and a microphone m via each of the sound pickups. ~Ιμ and convolution of the sound recorded in the sound field to estimate the direction of the sound source. Use the flowchart of Figure 9 to illustrate this -19-13880 processing. As shown in FIG. 9, generally, the direction estimating unit 22 first performs rotation of the directional beam by switching the filter coefficients prepared in advance (refer to the schematic diagram of FIG. 6), and measures in an arbitrary direction. The energy of the sound field (STEP1). Next, based on the measurement in STEP 1, the direction estimation algorithm is used to detect the direction of the sound source (STEP 2 ). Specifically, the direction estimating unit 22 uses an arbitrary forgetting factor to observe the energy in each direction, and estimates the existence direction of the sound source by the threshold of the set sound source. Next, it is confirmed whether or not the presence direction of the sound source is detected (STEP3). When the direction in which the sound source is present is not detected (NO), the process returns to STEP2, and the processing of STEP2 to STEP3 is repeated. On the other hand, when the direction in which the sound source is present is detected (YES), the process proceeds to STEP4. If the direction of the sound source is detected (YES of STEP 3), the filter coefficient group (STEP4) that emphasizes the direction is selected, and this is input to the directivity control unit 21 (STEP5). (3) The process of the directivity control unit is based on the process of the general direction estimating unit 22 described above, and the directivity control unit 21 inputs the sound corresponding to the direction of the directivity control unit 21 to emphasize the sound, and then the directivity control unit In 21, based on the direction and position of the input (distance from the radio device), the expected response function and the transfer function are selected in advance (or are calculated from the actual measurement and are taken -20 - 1380704). Control the point, and call out the expected response function matrix of the control point q and the transfer function matrix, and then calculate and calculate the control filter Hu~HMN by substituting the above into the above formula 3 . At this time, since the distance or direction between the respective sound collecting devices Ι1 to Im and the control point q are different, the expected response function and the transmission function are different, and when the reproduced channel system is plural In the case of the directivity of the sound receiving device in each channel (the direction emphasized by the sound collecting device and the sound receiving direction) is also different, the control filters are also different. In this way, if the control filters are determined, the control filters Hh to HMN are used, and in each of the channels, only the sound of the desired direction is used in the sound of each of the sound pickup devices. Emphasize. Then, the signal from each control filter is added to each channel through the adder ΣΝ!~ΣΝ, and these outputs the output unit 0 from the output of each channel. It is output to the reproduction device of each channel. Next, in the present embodiment, in order to perform the monitoring of the reproduced channel, the channel to be monitored is designated from the channel specifying unit 32. In this way, from the signals from the adders of the channels provided in the digital signal processing unit, only the signals of the desired channels are selected, and the signals are passed through the control filters S1, Cl~. Sn ' Cn ' is output to a 2-channel speaker or earphone as a playback device for monitoring. In this case, the coefficient of the control filter S!, (^~3n, Cn) is set in response to the output reproducing device, whereby the most appropriate output can be obtained regardless of the type of the reproducing device. -21 - 1380704 [4. Effect of the embodiment] In the above-described general embodiment, the direction estimating unit 22 uses a filter coefficient group that is a coefficient for forming directivity in each direction. The sound collection microphone h~IM is convoluted from the sound recorded in the sound field, and the sound pressure distribution in each direction is calculated using the direction estimation algorithm, and the energy is observed in each direction through an arbitrary forgetting factor. And the estimation of the existence direction of the sound source can be performed by the threshold value of the set sound source determination, and then the direction estimation information obtained by the direction estimation unit 22 is used to automatically exist for the sound source. In addition, even if it is reproduced at a long distance, it is possible to reconstruct the sound field which is collected using the estimation result of the direction, and to the binaural By using such a direction estimating unit, for example, it is possible to estimate the sound source direction of the robot, that is, to use the sensor to estimate the effect of the cocktail in the direction of the sound source in the robot ear. (Cocktail party, effect), etc. Also, it can be applied to a multi-channel recording system, that is, it can be applied to 5.1 ch recording, etc. Further, it can be used in a high-presence communication system, for example, can be utilized in a robot. In the conference system, etc., even if you do not go to the place, you can create the same sound field space as the place at a long distance. Also, because of the control point q set in advance, the expectation is made. The response and the transfer function are measured or calculated based on the measured 値 and the -22- 1380704 is calculated and the control filter is determined based on the data based on the actual measurement. Therefore, whether it is any of the radio devices When the direction is imparted to the directivity, the inverse matrix [(:(ω)τ·(:(ω) ]·丨C) of the transfer function matrix C constituting the control filter 均 can be used. (ω) τ is solved by an approximation calculation method such as the least square method, and an output approximate to the desired response is obtained. Further, in the present embodiment, the output is guided from the digital signal processing unit 2 Since the monitor processing unit 3 inputs and outputs the playback device to the two channels, the output of the playback channel is set to the channel specifying unit 32 provided in the monitor processing unit 3, regardless of the output of the playback channel. 'You can make a clear distinction and listen to other channel sounds. Of course, in this case, although you can only monitor the sound of a single reproduction channel, you can also use the adder!~ΣΝ The sound of the plurality of channels outputted is simultaneously outputted to the monitoring device. [Schematic Description] [Fig. 1] A diagram showing a configuration example of a microphone used in the present invention, (A) is a side view Figure, (B) is a front view. Fig. 2 is a diagram showing a reproduction equalization circuit for obtaining an algorithm for controlling a filter Η constituting the sound pickup system of the present invention. Fig. 3 is a view showing a state in which the desired response is set in the sound field space in the present invention. Fig. 4 is a block diagram showing one embodiment of the sound pickup system of the present invention. -23- 1380704 [Fig. 5] A diagram showing a state in which directivity is set in five directions around the microphone. Fig. 6 is a view showing a state in which the directivity due to the filter coefficient group is formed in the present invention. Fig. 7 is a block diagram showing one of the embodiments for generating an arbitrary forgetting factor in the present invention. Fig. 8 is a view showing a state of rotation of the number of directivity caused by switching the filter coefficient group in the present invention. [Fig. 9] A flowchart showing the processing of the direction estimating unit in the present invention. Fig. 10 is a block diagram showing an example of a conventional sound pickup system. [Main component symbol description]
Ml〜M4 :麥克風 1 :收音裝置 2 :數位訊號處理部 2 1 :指向性控制部 22 :方向推定部 3:監聽處理部 31 :假想音源再生處理部 3 2 :聲道指定部 4 :再生處理部 A :期望響應 C :傳達函數 -24- 1380704M1 to M4: microphone 1 : sound pickup device 2 : digital signal processing unit 2 1 : directivity control unit 22 : direction estimation unit 3 : monitoring processing unit 31 : virtual sound source reproduction processing unit 3 2 : channel designation unit 4 : reproduction processing Part A: Expected Response C: Communication Function - 24 - 1380704
Η :控制濾波器 h-lM:收音用麥克風（收音用裝置） Η,-Ημ:收音系統用之控制濾波器 Σ 1〜Σ ν :加算器 〇ι〜On:再生輸出部 S!，Ci-Sn，Cn:假想音源再生處理部之控制濾波器 〇,，〇2 :監聽用輸出部 -25 -Η : Control filter h-lM: Radio microphone (receiving device) Η, -Ημ: Control filter for radio system Σ 1~Σ ν : Adder 〇ι~On: Regeneration output S!, Ci- Sn, Cn: Control filter for the virtual sound source reproduction processing unit, 〇2: Monitor output unit -25 -