JP3779554B2 - Frequency offset estimation method and frequency offset estimator - Google Patents

Description

ただし、Ｌ_c ＋１は伝送路の最大遅延量、ａ_i,l ；ｉ＝０,…,Ｌ_c は１番目の干渉波のインパルス応答、ω_l とθ_0,l はｌ番目の信号の周波数と初期位相、Ｔはシンボル周期、ｎ_k は雑音成分である。
【００１９】
実際、干渉という場合には様々なシステムからの干渉が含まれるが、ここでは簡単のため、同じシステムからの干渉とし、さらに（１）式では各々が同期をとっているＴＤＭＡに近いシステムを想定している。この時、ｉ＝１番目の信号を所望信号と仮定する。すなわち、受信信号に含まれる信号のうち、ｉ＝１番目の信号の周波数成分を推定することを考える。ここでまず、受信信号にｉ＝１番目の既知の信号パターンを乗算すると以下の信号が得られる。
【００２０】
【数２】

ここで、ｙ_k の時間系列を要素とするＬ次元のベクトルＹ_k ＝［ｙ_k ｙ_k-1 … ｙ_k-L+1 ］^Tを定義する。すると、ベクトルＹ_k は以下のように書き換えられる。その相関行列は以下のように表される。
【００２１】
【数３】

上の（３．１）〜（３．６）式における添え字ＴとＨは各々ベクトルあるいは行列の転置とエルミート転置である。この時、Ｙ_k の相関行列Ｒ_k は以下のように表される。
【００２２】
【数４】

ただし、Ｅ［…］は集合平均をとることを意味しており、Ω_0,l とＣ_l は以下のように表される。
【００２３】
【数５】

ただし、（４．１）式と（４．３）式の導出では雑音と信号の無相関性、異なったユーザの信号間の無相関性を利用した。また、（４．１）式のＩは単位行列、σ² は雑音の分散である。
【００２４】
ここで、（４．１）式のＥ［Ｃ］＝｛Ｅ［ｃ_i,j ］｝は実際には次のようになる。
【００２５】
【数６】

上の（５．１）式と（５．２）式を（４．１）式に代入すると次式が得られる。
【００２６】
【数７】

ここで、（６．１）式を以下のように変形する。
【００２７】
【数８】

上記の相関行列Ｒ_k を固有値分解すると、その固有値λ_i は以下のようになる。
【００２８】
【数９】

（８．１）式と（８．２）式は（７．１）式が以下のように変形できることを意味している。
【００２９】
【数１０】

（８．１）式と（８．２）式よりλ₁ ＞λ₂ であるため、相関行列の固有値の最大値に対応する固有ベクトルは搬送波ベクトル、あるいは周波数オフセットベクトルになっている。従って、相関行列の最大固有値に対応した固有ベクトルの周波数を測定することにより、周波数が推定できる。また、（８．１）式と（８．２）式を見ても分かるように、固有値分解を利用することで所望信号のコヒーレント成分を含む固有値λ₁ とコヒーレント成分を含まない固有値λ₂ とに分離できるため、高精度な周波数推定が可能になることが分かる。
【００３０】
以上の原理的考察のもとに、本発明は次の点を特徴とする。
【００３１】
請求項１の発明は、受信信号の搬送波周波数を推定し、推定結果に基づき搬送波帯からベースバンド帯に変換し、信号を復調する放送あるいは通信用の受信方法における周波数オフセット推定方法であって、前記受信信号に既知の信号を掛け合わせる乗算処理と、前記乗算処理の結果出力を所定時刻間隔で出力する遅延処理と、前記遅延処理の出力を要素とするベクトルの相関行列を演算する相関行列演算処理と、前記相関行列演算処理の結果出力である相関行列の固有値を求め、最大固有値に対応する固有ベクトルを搬送波の波形として出力する固有値分解処理とを有するものである。
【００３２】
請求項２の発明は、受信信号の搬送波周波数を推定し、推定結果に基づき搬送波帯からベースバンド帯に変換し、信号を復調する放送あるいは通信用の受信方法における周波数オフセット推定方法であって、前記受信信号に既知の信号を掛け合わせる乗算処理と、前記乗算処理の結果出力を所定時刻間隔で出力する遅延処理と、前記遅延処理の出力を要素とするベクトルの相関行列を演算する相関行列演算処理と、前記相関行列演算処理の結果出力である相関行列の固有値を求め、最大固有値に対応する固有ベクトルに含まれる周波数成分の最大値を推定して出力する固有ベクトルの周波数推定処理とを有するものである。
【００３３】
請求項３の発明は、受信信号の搬送波周波数を推定し、推定結果に基づき搬送波帯からベースバンド帯に変換し、信号を復調する放送あるいは通信用の受信設備における周波数オフセット推定器であって、前記受信信号に既知の信号を掛け合わせる乗算手段と、前記乗算手段の出力を所定時刻間隔で出力する遅延手段と、前記遅延手段の出力を要素とするベクトルの相関行列を演算する相関行列演算手段と、前記相関行列演算手段の出力である相関行列の固有値を求め、最大固有値に対応する固有ベクトルを搬送波の波形として出力する固有値分解手段とを備えて成るものである。
【００３４】
請求項４の発明は、受信信号の搬送波周波数を推定し、推定結果に基づき搬送波帯からベースバンド帯に変換し、信号を復調する放送あるいは通信用の受信設備における周波数オフセット推定器であって、前記受信信号に既知の信号を掛け合わせる乗算手段と、前記乗算手段の出力を所定時刻間隔で出力する遅延手段と、前記遅延手段の出力を要素とするベクトルの相関行列を演算する相関行列演算手段と、前記相関行列演算手段の出力である相関行列の固有値を求め、最大固有値に対応する固有ベクトルに含まれる周波数成分の最大値を推定して出力する固有ベクトルの周波数推定手段とを備えて成るものである。
【００３５】
【発明の実施の形態】
以下、本発明の実施の形態を図に基づいて詳説する。
【００３６】
＜第１の実施の形態＞
図１は、本発明の第１の実施の形態の周波数オフセット推定器の構成を示す。同図において、１００は入力端子、１０１は乗算器、１０２はパイロット信号記憶素子（pilot symbols）、１０３は遅延回路、１０４は固有値分解部、１０５は相関行列演算部、１０６は推定周波数の出力端子である。同図では、入力端子１００の入力信号はパイロット信号記憶素子１０２から出力される既知信号と掛け合わされ、（２）式のｙ_k を得る。このｙ_k を遅延回路１０３に入力すると、遅延回路１０３からは各々遅延時間の異なる信号群ｙ_k ,ｙ_k-1 ,…,ｙ_k-L+1 が出力される。この信号群はベクトルＹ_k の要素に等しい。
【００３７】
そこで相関行列演算部１０５で、このベクトルＹ_k を用いて以下の演算を行うことで相関行列Ｒ_k を得る。
【００３８】
【数１１】

（４．１）式では相関行列はＹ_k Ｙ_k ^H の集合平均で求めていたが、実際にはエルゴードの定理に基づき時間平均で求めている。この相関行列は（１１）式の演算式からも分かるようにエルミート行列である。従って、その固有値分解はJacobi法やＱＲ法等、様々な手法で求めることが可能である。固有値分解部１０４は、この固有値分解処理を行う。（黒瀬；「Fortran９０のためのサブルーチンライブラリ〜数値計算、統計計算、３次元ＣＧライブラリ〜」、森北出版、１９９８年）。
【００３９】
固有値分解部１０４は、こうして得られた固有値および固有ベクトルのうち、最大固有値に対応する固有ベクトルだけを取り出し、これを周波数オフセットの推定値として出力端子１０６から出力する。
【００４０】
上記の周波数オフセット推定器における遅延回路１０３の詳しい構成を、図２に示す。同図において、２０１は入力端子、２０２〜２０４は遅延素子、２０５〜２０８は信号群の出力端子を表している。入力端子２０１からｙ_k が入力されると、これに遅延をかけて、ｙ_k ,ｙ_k-1 ,…,ｙ_k-L+1 の信号を出力端子２０５〜２０８から出力する。
【００４１】
＜第２の実施の形態＞
次に、本発明の第２の実施の形態の周波数オフセット推定器を、図３に示す。同図において、１１１は入力端子、１１２は乗算器、１１３は遅延回路、１１４はパイロット信号記憶素子、１１５は相関行列演算部、１１６は固有値分解部、１１７は固有ベクトルの周波数推定部、１１８は推定周波数の出力端子である。
【００４２】
第２の実施の形態では、図１に示した第１の実施の形態と同様に受信信号ベクトルの相関行列から固有値を求め、その最大固有値に対応したベクトルを生成する。この生成されたベクトルは必ずしも正弦波とはならないので、この固有ベクトルのもつ最も高い周波数成分を固有ベクトルの周波数推定部１１７によって求め、周波数オフセットの推定値として出力端子１１８より出力する。
【００４３】
図４に、上記の固有ベクトルの周波数推定部１１７の詳しい構成を示す。同図において、２２１−１〜２２１−４は固有値分解部１１６からの固有ベクトルの各要素の入力端子、２２２−１〜２２２−４は固有ベクトルを記憶しておくバッファ回路、２２３〜２２９は乗算器、２３０は固定値「１」を入力する端子、２３１は加算器、２３４は適応制御部、２３５は出力端子である。また、適応制御部２３４では固有ベクトルｖ₁ ＝［ｖ_1,0 ,ｖ_1,1 ,…,ｖ_1,L-1 ］に対して以下の演算を繰り返し行う。
【００４４】
【数１２】

ただし、（１２．３）式のμはステップサイズパラメータと呼ばれる微小係数、Im［…］は虚数部だけを抽出する関数を表している。
【００４５】
上記の演算を所定の回数ｍだけ繰り返した後、
【数１３】

をもって推定周波数として端子２３５から出力する。
【００４６】
＜第３の実施の形態＞
本発明の第３の実施の形態の周波数オフセット推定器を、図５に示す。一般に無線システムでは、既知の無線周波数帯に信号が配置されているため、おおよその周波数は分かっている。ところが、場合によっては、装置の不完全性などにより想定した初期周波数より大きくずれている場合がある。第３の実施の形態の周波数オフセット推定回路は、このように推定周波数が推定の初期値より大きくずれている場合に有効な回路である。
【００４７】
同図において、１２１は入力端子、１２２は乗算器、１２３はパイロット信号記憶素子、１２４は遅延回路、１２５は相関行列演算部、１２６は固有値分解部、１２７は固有ベクトルの周波数推定部、１２８はＤＦＴ回路、１２９は周波数オフセットの推定値の出力端子、１３０は最大値検出部である。
【００４８】
第３の実施の形態の回路は、図３に示した第２の実施の形態の回路と同様の処理を行うが、一点異なるのは最大固有値に対応した固有ベクトルをＤＦＴ回路１２８に入力してＤＦＴ処理を行い、最大値検出部１３０によりその最大値に相当する周波数成分を固有ベクトルの周波数推定における初期値、
【数１４】

として与え、固有ベクトルの周波数推定部１２７はこれを用いて周波数推定を行う点である。
【００４９】
原理的に、第２の実施の形態において採用した図４に示す周波数推定部１１７による固有ベクトルの周波数推定は高精度であるが、あまり初期値が大きくて実際の周波数と異なる場合には、全く異なった値を出力することがある。そこで、第３の実施の形態では、ＤＦＴ回路１２８により粗く推定された周波数を初期値として用いることで、この点を克服することができる。
【００５０】
＜適用例１＞
図６に本発明の周波数オフセット推定器を用いた受信機を示す。同図において、１３１は搬送波帯信号入力端子、１３２は局部発振器、１３３〜１３５は乗算器、１３６は遅延素子、１３７は図３に示した第２の実施の形態の周波数オフセット推定器あるいは図５に示した第３の実施の形態の周波数オフセット推定器、１３８は信号出力端子、１３９は帯域通過フィルタである。
【００５１】
この受信機では、受信した信号を一旦ベースバンド帯域近くまで周波数変換し、そのときのＩＦ周波数である局部発振器と受信信号の搬送波周波数の誤差である周波数オフセットを推定するである。すなわち、入力端子１３１から入力された受信信号を局部発振器１３２からの信号と乗算器１３３で掛け合わせ、ＩＦ周波数帯の信号とする。この信号を、帯域通過フィルタ１３９を経た後に分岐し、片方は周波数オフセット推定器１３７に入力し、周波数オフセットを推定する。そして、推定周波数オフセットに対して乗算器１３５と遅延素子１３６とで位相の積分処理を行い、周波数オフセットによる位相誤差を乗算器１３５より出力し、これと分岐された帯域通過フィルタ１３９の出力信号と乗算器１３４によって掛け合わせることで完全なべ一スバンド帯の信号を得、信号出力端子１３８より出力する。
【００５２】
＜適用例２＞
図７に本発明の周波数オフセット推定器を採用したＣＤＭＡ（Code Division Multiple Access）用基地局の受信設備を示す。同図において、１４０は帯域通過フィルタ、１４１は受信信号入力端子、１４２は局部発振器、１４３〜１４５は乗算器、１４６は遅延素子、１４７は周波数オフセット推定器、１４８は外部パイロット信号記憶回路、１４９〜１５１は周波数オフセット補償部（ＡＦＣ）、１５２〜１５４はＣＤＭＡの受信機、１５５〜１５７は復調信号出力端子を表している。ＣＤＭＡでは数人のユーザが同時に同じ無線帯域を利用して通信を行う。従って、受信信号には数人のユーザの信号が混入している。さらにその無線搬送波周波数も厳密にいえば、各ユーザごとに少しずつ異なっている。図７の受信設備はこれらの全てのユーザの信号を復調する構成である。
【００５３】
この第２の適用例の場合、図６に示した第１の適用例と同様に、局部発振器１４２によりＩＦ周波数帯にまで変換された受信信号を、＃１〜＃Ｎの分岐され周波数オフセット補償器１４９〜１５１に入力する。
【００５４】
周波数オフセット補償部１４９〜１５１の各々におけるパイロット信号記憶素子には、外部パイロット信号記憶回路１４８より各々異なったパイロット信号がダウンロードされる。そこで、周波数オフセット補償部１４９〜１５１の各々では、周波数オフセット推定器１４７と乗算器１４４，１４５と遅延素子１４６により、図６で説明したように受信信号のＩＦ周波数を除去する。ここで注意すべきことは、周波数オフセット補償器１４９〜１５１ではパイロット信号記憶回路１４８がダウンロードしたパターンを信号内に含む信号、すなわち所望信号だけがベースバンドに変換され、それ以外の信号はベースバンドに変換されるという保証はない。周波数オフセット補償器１４９〜１５１の各出力は受信機１５２〜１５４各々に入力され、復調信号を得る。
【００５５】
図８に図７のＣＤＭＡ受信設備で利用したＣＤＭＡ受信機１５２〜１５４の構成を示す。同図において、１６１は入力端子、１６２は遅延回路、１６３〜１６５は逆拡散用のマッチドフィルタ、１６６〜１６８は乗算器、１６９は加算器、１７０は適応制御部、１７１は復調信号出力端子である。この受信機の構成は、いわゆるRake受信機と呼ばれる構成であり、先行波に加えて遅延波も同相加算することでＳＮＲ（信号ノイズ比）を最大化することが可能である。
【００５６】
適応制御部１７０の構成としては、周波数オフセットが既に除去されているため、線形推定理論に則ったアルゴリズムであるＬＭＳ（Least Mean Square）法やＲＬＳ（Recursive Least Squares）法等が適用できる。以下には、一例としてＬＭＳ法を適用した場合のアルゴリズムを示す。まず、マッチドフィルタ１６３〜１６５の各出力をＺ_k ＝［ｚ_k,1 ｚ_k,2 … ｚ_k,M ］^T とし、適応制御部１７０からの係数ベクトルをＷ_k ＝［ｗ_k,1 ｗ_k,2 … ｗ_k,M ］^T とする。ただし、Ｍはマッチドフィルタの数である。その時、
【数１５】

というアルゴリズムにより復調信号を得ることができる。ただし、Ｄ_k はＷ_k ^H Ｚ_k の判別値、あるいはトレーニング信号と呼ぶ、既知のシンボルパターンである。
【００５７】
図９に、図８のマッチドフィルタ１６３〜１６５の具体的な回路構成を示す。同図は拡散率が４の場合の構成例である。同図において、２４０は入力端子、２４１〜２４３は遅延素子、２４４〜２４７は拡散符号入力端子、２４８〜２５１は乗算器、２５２は加算器、２５３は出力端子である。
【００５８】
＜適用例３＞
図１０に本発明の周波数オフセット推定器を適用したマルチユーザ受信設備の構成を示す。同図では、１８１は受信信号入力端子、１８２は局部発振器、１８３は乗算器、１８４は帯域通過フィルタ、１８５は周波数オフセット推定器、１８６は外部パイロット信号記憶回路、１８７〜１９０はパイロット信号ダウンロード機能つき周波数オフセット推定器、１９１はマルチユーザ受信機、１９２〜１９４は復調信号出力端子を表している。
【００５９】
この構成のマルチユーザ受信設備では、受信信号に含まれる信号の周波数オフセットをパルス信号ダウンロード機能つき周波数オフセット推定器１８７〜１９０で推定する。各ユーザの周波数オフセットの推定値はマルチユーザ受信機１９１に入力する。マルチユーザ受信機１９１では、各ユーザの周波数オフセットの推定値やトレーニング系列と呼ばれる情報系列の間に埋め込まれた既知信号を利用して各ユーザの信号を一括して復調する。その復調信号は出力端子１９２〜１９４から出力する。
【００６０】
図１１に、図１０に示したマルチユーザ受信機１９１の第１の構成例を示す。同図は４ユーザ受信機の構成例である。同図において、２６１は入力端子、２６２〜２６５は推定されたオフセット周波数の入力端子、２６６は減算器、２６７は二乗回路、２６８は最尤推定器（ＭＬＥ：Maximum Likelihood Estimator）、２６９〜２７２はレプリカ（replica）生成器、２７３は各ユーザの伝送路インパルス応答を推定する適応制御部、２７４〜２７７は復調信号出力端子である。同図は通常のマルチユーザ受信機に、周波数オフセット情報を加味してレプリカを生成して、最も尤度の高い復調信号を得るものである。
【００６１】
図１２に、図１１のマルチユーザ受信機で採用したレプリカ生成器の具体的な構成を示す。同図において、２８１は入力端子、２８２は遅延回路、２８３〜２８８は乗算器、２８９は図１１の適応制御部２７３からのインパルス応答入力端子、２９０は加算器、２９１は遅延素子、２９２はオフセット周波数入力端子、２９３はレプリカ出力端子である。
【００６２】
このレプリカ生成器では、入力端子２８１からの仮判定値を遅延回路２８２と乗算器２８３〜２８６とで構成されたＦＩＲ（Finite Impulse Response）フィルタによって畳み込み、ベースバンドレプリカを得る。一方、オフセット周波数入力端子２９２からのオフセット周波数信号を乗算器２８８と遅延素子２９１とで積分することで、オフセット周波数による位相変動成分を演算する。そして乗算器２８７により、この位相変動成分にベースバンドレプリカを乗算することで受信信号のレプリカを生成し、出力端子２９３から図１１のＭＬＥ回路２６８に出力する。
【００６３】
図１３に、図１１のマルチユーザ受信機１９１に適用するＭＬＥ回路２６８の構成例を示す。同図はＢＰＳＫ変調信号を用いた４ユーザに対する最尤推定回路である。同図において、３０１は図１１の二乗回路２６７よりの信号入力端子、３０２はリセット信号入力端子、３０３〜３０７はスイッチ回路、３０８は減算器、３０９は判別器、３１０は遅延素子、３１１はバイナリカウンタ、３１２はクロック入力端子、３１３，３１４，３１９，３２０は復調信号出力端子、３１５〜３１８は仮判定値出力端子である。
【００６４】
このＭＬＥ回路２６８では、バイナリカウンタ３１１をクロックにより駆動して、ＢＰＳＫ変調を受けた４ユーザの信号全てを発生させる。発生させた仮判定値に基づき４ユーザのレプリカを生成し、レプリカと受信信号の二乗ユークリッド距離をスイッチ回路３０４に入力させる。減算器３０８と遅延素子３１０、スイッチ回路３０３から構成される最低値検出器によって二乗ユークリッド距離が最低の仮判定値を求め、それをスイッチ回路３０３により選択して復調信号として出力する。
【００６５】
図１１に示したマルチユーザ受信機１９１における適応制御部２７３も、図７のＣＤＭＡ用受信機１５２〜１５４と同様に、各ユーザの周波数オフセットが既知なので、ＬＭＳやＲＬＳといった線形推定理論に基づくアルゴリズムが適用できる。
【００６６】
図１４に、図１０に示したマルチユーザ受信機１９１の第２の構成例を示す。これは、複数のアンテナ素子で受信し、これをＫ分岐してＫ個の各分岐先で適応的にビームを形成するマルチビームアンテナを備えている。同図において、３３０は図１０における周波数オフセット推定器によって推定されたオフセット周波数の入力端子、３３１はＮ個のアンテナ素子で受信された受信信号の入力端子、３３２〜３３４はビーム形成器（ＢＦＮ：Beam Forming Network）、３３５〜３３７は復調信号出力端子である。
【００６７】
この構成のマルチユーザ受信機では、ｉ番目のビーム形成器では、１〜ｉ−１番目までのユーザの信号をキャンセルし、その後に出力信号のＳＩＮＲ（Signal to Interference and Noise Ratio）が最大になるようアンテナの指向性を制御する。この時、信号を正確にキャンセルするために、推定された周波数オフセットに基づき各復調信号に位相回転を与えることで、その制御アルゴリズムに線形推定理論に基づくアルゴリズムを用いることができ、高い干渉抑圧特性を達成できる。実際の通信路では、各ユーザごとに周波数オフセットが異なるため、図１４に示す構成を用いて干渉キャンセルを行うことで、現実の通信路でも高い通信品質を達成することが可能になる。
【００６８】
図１５に、図１４のマルチユーザ受信機１９１におけるｊ番目のビーム形成器ＢＮＦ（j）の構成例を示す。同図において、３４１〜３４４は各アンテナ素子からの信号入力端子、３４５〜３５１，３６０，３６２，３６７〜３６９は乗算器、３５２は加算器、３５３は減算器、３５４は判定器、３５５はスイッチ回路、３５６はトレーニング信号と呼ばれる既知信号入力端子、３５７は復調信号出力端子、３５８は適応制御部（Adaptive Weight Control）、３７１〜３７３はｊ−１番目のＢＦＮ（ｊ−１）が復調した信号の入力端子、３６４は推定された周波数オフセットの入力端子、３６１は遅延素子、３６３は積分器、３６５は複素共役（・）^* を表している。
【００６９】
適応制御部３５８では、復調信号出力端子３５７からの出力信号を所望信号、減算器３５３の出力信号を誤差信号、乗算器３６７〜３６９の出力および入力端子３４１〜３４４よりの信号を入力信号とする線形推定アルゴリズム、例えばＬＭＳやＲＬＳアルゴリズムを用いて乗算器３４５〜３５１の係数を適応的に推定する。
【００７０】
図１６に従来技術であるＲＬＳ適応位相制御法による周波数オフセット推定誤差特性を示す。推定誤差の二乗平均を縦軸にし、横軸は受信シンボル数を表している。同図では、ＡＷＧＮチャネルと２パスモデル（ＩＳＩ）、１波同一チャネル干渉（ＣＣＩ）、１波同一チャネル干渉に加えて各チャネルが２パスモデルである場合（ＣＣＩ＋ＩＳＩ）の特性を示す。同図では、ＣＮＲ＝０ｄＢ、周波数オフセット△fT＝１０^-3の条件で特性を取得した。ただし、Ｔはシンボル周期を表している。この特性から分かるように、ＲＬＳ適応位相制御法では、ＡＷＧＮチャネルにおいては高精度な推定が可能であるが、干渉が入った途端に劣化が始まり、ＣＣＩ＋ＩＳＩ伝送路では１０^-2程度の推定精度しか得られていない。
【００７１】
図１７に本発明の特性例を示す。条件は図１６に示した従来例の特性取得の場合と同様にした。１４０シンボルのトレーニング信号だけを用いて特性を評価した。周波数オフセットΔfT＝１０^-3の場合、１４０シンボルのトレーニング信号区間では位相が２π回転しないために若干の推定誤差が見られるが、推定誤差は３×１０^-4までに追い込む優れた特性であることが分かる。
【００７２】
【発明の効果】
本発明によれば、厳しい同一チャネル干渉・符号間干渉環境下において高精度に周波数オフセット推定が可能になる。この高精度な周波数オフセット情報を用いることで、移動通信システムにおける基地局の受信機の通信品質を高めるだけでなく、強力な干渉補償機能をもつアダプティブアレーやマルチユーザ受信機等を現実の通信に適用できるようにする。
【００７３】
すなわち、線形推定理論に基づくチャネル推定を必須とするマルチユーザ受信機を実際の通信に用いるには、劣悪な環境条件においても、各ユーザごとに異なる周波数オフセットを高精度に推定する周波数オフセット推定器が必須であったが、本発明はこの問題を解決することで、周波数利用効率の大幅な増大を可能とするマルチユーザ受信機を商用システムに導入することを可能にし、周波数利用効率の向上に資することができる。
【図面の簡単な説明】
【図１】本発明の第１の実施の形態の周波数オフセット推定器のブロック図。
【図２】上記の実施の形態における遅延回路のブロック図。
【図３】本発明の第２の実施の形態の周波数オフセット推定器のブロック図。
【図４】上記の実施の形態における固有ベクトルの周波数推定部のブロック図。
【図５】本発明の第３の実施の形態の周波数オフセット推定器のブロック図。
【図６】本発明の第４の実施の形態の受信設備のブロック図。
【図７】本発明の第５の実施の形態の受信設備のブロック図。
【図８】上記の第４及び第５の実施の形態における受信機のブロック図。
【図９】上記の受信機におけるマッチドフィルタのブロック図。
【図１０】本発明の第６の実施の形態の受信設備のブロック図。
【図１１】上記の第６の実施の形態におけるマルチユーザ受信機のブロック図。
【図１２】上記のマルチユーザ受信機におけるレプリカ生成器のブロック図。
【図１３】上記のマルチユーザ受信機における最尤推定回路（ＭＬＥ）のブロック図。
【図１４】上記の第６の実施の形態におけるマルチユーザ受信機の他の構成例を示すブロック図。
【図１５】上記のマルチユーザ受信機におけるｊ番目のビーム形成器ＢＦＮ（ｊ）のブロック図。
【図１６】従来例の特性を示すグラフ。
【図１７】本発明の特性を示すグラフ。
【図１８】従来のコスタス型搬送波再生器のブロック図。
【図１９】従来の直交検波器のブロック図。
【図２０】従来の逆変換型搬送波再生器のブロック図。
【図２１】従来の周波数オフセット補償機能つき遅延検波器のブロック図。
【図２２】従来のＲＬＳ位相制御回路のブロック図。
【図２３】従来のＤＦＴを適用した周波数オフセット推定器のブロック図。
【符号の説明】
１０２ パイロット信号記憶素子
１０３ 遅延回路
１０４ 固有値分解部
１０５ 相関行列演算部
１１４ パイロット信号記憶素子
１１３ 遅延回路
１１５ 相関行列演算部
１１６ 固有値分解部
１１７ 固有ベクトルの周波数推定部
１２３ パイロット信号記憶素子
１２４ 遅延回路
１２５ 相関行列演算部
１２６ 固有値分解部
１２７ 固有ベクトルの周波数推定部
１２８ ＤＦＴ
１３０ 最大値検出部[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a frequency offset estimation method and a frequency offset estimator.
[0002]
[Prior art]
In many communication systems and broadcasting systems, a transmission signal is once converted into a frequency in a carrier band, and the signal is transmitted. On the receiver side, signal transmission is performed by accurately removing a carrier frequency component from the reception signal. At this time, in many systems, the carrier frequency is estimated from the received signal. In general, this carrier frequency estimation accuracy greatly affects signal transmission characteristics. In particular, in a digital signal transmission system to which phase modulation is applied, information is put on the phase of a carrier wave to be transmitted. In order to extract this information by a receiver, accurate carrier frequency estimation is essential.
[0003]
Conventionally, in a wired communication system using a metallic cable, a fixed wireless system or a satellite communication system with stable transmission characteristics, this carrier estimator and demodulator are integrated into a Costas type carrier regenerator so that only the carrier frequency can be obtained. However, many methods have been used to estimate the phase simultaneously and with high accuracy. FIG. 18 shows a configuration example of a Costas type carrier recovery circuit. In the figure, 1 is a received signal input terminal, 2 is a quadrature detector, 3 is a voltage controlled oscillator, 4 to 6 are multipliers, 7 is a subtractor, 8 is an adder, 9 is a loop filter, 10 And 11 are in-phase / quadrature signal output terminals. FIG. 19 shows a detailed configuration of the quadrature detector 2 used in FIG. In the figure, 12 is an input terminal, 13 and 17 are branch circuits, 14 and 15 are multipliers, 16 is a π / 2 phase shifter, and 19 and 20 are output terminals for in-phase and quadrature signals.
[0004]
In addition to the Costas type, an inverse modulation type carrier recovery circuit is known. In this method, a received signal is demodulated in advance, a carrier wave component is extracted, the frequency of an oscillator is synchronized with the carrier wave, and the carrier wave is regenerated, and this demodulated carrier wave regeneration circuit is a TDMA (Time Division Multiple Access). ) Method has been used in satellite communications. FIG. 20 shows the configuration of an inverse modulation type carrier recovery circuit. In the figure, 21 is a received signal input terminal, 22 is a preamble signal input terminal, 23 and 24 are branch circuits, 25 to 27 are multipliers, 28 is a band pass filter called a tank filter, and 29 is a loop filter. (Loop filter), 30 is a voltage controlled oscillator, and 31 is an output terminal.
[0005]
On the other hand, in a mobile communication system in which transmission path fluctuations are severe, a delay detection method capable of high-speed phase synchronization is used. In a receiver using delay detection, the frequency is once converted to near baseband by a local oscillator close to the carrier frequency, and this received signal is subjected to delay detection. At this time, a stationary phase error caused by the oscillation frequency of the local oscillator and the carrier frequency of the received signal appears in the signal after delay detection. By correcting this steady phase error, the carrier frequency is estimated equivalently.
[0006]
FIG. 21 shows a configuration of a delay detector having a frequency difference (referred to as “frequency offset”) compensation function between the local oscillator and the carrier frequency. In the figure, 32 is a received signal input terminal, 33 to 36 are multipliers, 37 is an adder, 38 is a subtractor, 39 and 40 are 1-symbol delay elements, 41 is a discriminator, and 42 is a coefficient input of the multiplier 36. 43, a delay detector, 44 a frequency offset estimation unit, 45 a local oscillator, and 46 a demodulated signal output terminal.
[0007]
In principle, the above-described delayed detection method has a characteristic deterioration of slightly less than 3 dB as compared with a Costas type synchronous detection method. For this reason, studies have been made to apply a synchronous detection method that is advantageous in characteristics to a mobile communication system. Among these, the phase control method to which the RLS (Recursive Least Squares) algorithm is applied is intended for TDMA mobile communication systems such as PDC (Personal Digital Cellular) and PHS (Personal Handy-phone System), at the head of each TDMA slot. Not only phase synchronization but also frequency synchronization can be established at high speed. Moreover, since this method can be realized with a relatively small circuit scale, it can be said to be an excellent method for application to a mobile communication system that requires a reduction in size and power consumption. FIG. 22 shows a configuration example of the RLS phase control circuit. In the figure, 47 is a received signal input terminal, 48 and 49 are multipliers, 50 and 51 are discriminators, 52 is a one-symbol delay element, 53 is a terminal for inputting a fixed value “1”, and 54 and 55 are RLS controls. Reference numeral 56 denotes a demodulated signal output terminal.
[0008]
In a mobile communication system, the same channel is separated and arranged in order to improve surface frequency utilization efficiency. That is, several types of carrier frequencies f₁~ F_nAre prepared, and carrier frequencies different from the adjacent base stations are arranged. Therefore, by using the same frequency for every n base stations, it is possible to cover all service areas with n carrier frequencies, assuming that the base stations are arranged linearly. In this configuration, in order to improve frequency utilization efficiency, it is desirable to shorten this distance, that is, to reduce the frequency repetition number n. However, in that case, co-channel interference occurs.
[0009]
It is known that this co-channel interference can be removed by applying interference cancellation such as an adaptive array. However, when an adaptive array or the like is applied, since an algorithm for controlling these is generally based on linear estimation theory, it is necessary to simultaneously mount high-accuracy frequency offset estimation means. In other words, accurate frequency offset estimation is required in the presence of co-channel interference waves, but accurate demodulation is not possible with methods that involve demodulation, such as the Costas type carrier recovery method and inverse modulation type carrier recovery method. There is a problem that accurate frequency offset estimation cannot be performed.
[0010]
Therefore, a frequency offset estimation method using DFT (Digital Fourier Transform) has been proposed as a method of performing frequency offset estimation even under such interference conditions. In this method, a known pattern signal is inserted into a part of a transmission signal, and the receiver removes the known pattern signal component from the received signal and extracts a carrier wave component. The DFT process is performed on the extracted carrier wave, and the maximum value is set as the carrier frequency. This method enables accurate frequency offset estimation even in the presence of interference waves due to the excellent frequency analysis capability of DFT.
[0011]
FIG. 23 shows a configuration example of a frequency offset estimator using DFT. In the figure, 61 is a received signal input terminal, 62 is a local oscillator, 63 and 64 are multipliers, 65 is a known pattern signal (pilot signal) storage element, 66 is a DFT, 67 is a maximum value detector, Reference numeral 68 denotes an estimated frequency output terminal.
[0012]
This frequency offset estimation method using DFT has a considerably high frequency offset estimation capability as compared with a method involving demodulation, but the frequency offset estimation accuracy achieved by the carrier recovery circuit in AWGN (Additive White Gaussian Noise). In comparison, it is greatly degraded. Although it is an unrealistic model, accurate frequency offset estimation is difficult even if all are known pattern signals.
[0013]
By the way, in order to provide a comfortable multimedia service in a mobile communication environment, high-speed signal transmission is required even in a mobile radio transmission path. As is well known, when high-speed communication is performed on a mobile communication transmission line in which an object such as a house causing reflection / diffraction exists, the characteristics are significantly deteriorated due to severe intersymbol interference caused by multipath. An adaptive equalizer is effective to compensate for this. However, since the control algorithm is based on the linear estimation theory as in the adaptive array, it is necessary to provide a highly accurate frequency offset estimation circuit. Although a method for realizing a high-accuracy frequency offset estimator by combining an adaptive equalizer and a frequency offset estimation circuit is known, in addition to intersymbol interference caused by multipath, inter-channel interference is further reduced. In the existing environment, it is difficult to perform accurate frequency offset estimation as described above.
[0014]
[Problems to be solved by the invention]
As described above, conventionally, there has been a problem that it is difficult to accurately estimate frequency offset necessary for signal demodulation under a severe co-channel interference environment.
[0015]
Also, when providing a multimedia mobile communication system, intersymbol interference due to multipath occurs in addition to co-channel interference, but in a transmission path where this inter-symbol interference and co-channel interference occur However, the conventional frequency estimator provided in the demodulator for the Gaussian channel, Rayleigh fading, and intersymbol interference transmission path has a problem that it is difficult to estimate the frequency with high accuracy.
[0016]
The present invention has been made to solve such a conventional technical problem, and accurately estimates the frequency of a carrier wave in a receiving facility used in a broadcast or communication system that performs signal transmission by converting a signal into a carrier wave band. It aims at providing the technology to do.
[0017]
[Means for solving problems]
Transmit signal x_{k, l}Is the carrier angular frequency ω_lIs sent to the transmission line. Here, the subscript l represents the user number, and k represents the time. At this time, when the signals of 1 to K users interfere with each other on the transmission line, and there is a delayed wave in each, the received signal s_kCan be written as:
[0018]
[Expression 1]

However, L_c+1 is the maximum delay amount of the transmission line, a_{i, l}I = 0, ..., L_cIs the impulse response of the first interference wave, ω_lAnd θ_{0, l}Is the frequency and initial phase of the l-th signal, T is the symbol period, and n_kIs a noise component.
[0019]
Actually, interference includes interference from various systems, but here it is assumed that the interference is from the same system for the sake of simplicity. Furthermore, in Equation (1), a system close to TDMA, each of which is synchronized, is assumed. is doing. At this time, it is assumed that i = 1st signal is a desired signal. That is, let us consider estimating the frequency component of the i = 1st signal among the signals included in the received signal. Here, when the received signal is first multiplied by the i = 1th known signal pattern, the following signal is obtained.
[0020]
[Expression 2]

Where y_kL-dimensional vector Y whose elements are time series of_k= [Y_k y_k-1 ... y_{k-L + 1}]^TDefine Then the vector Y_kCan be rewritten as: The correlation matrix is expressed as follows.
[0021]
[Equation 3]

The subscripts T and H in the above equations (3.1) to (3.6) are vector or matrix transpose and Hermitian transpose, respectively. At this time, Y_kCorrelation matrix R_kIs expressed as follows.
[0022]
[Expression 4]

However, E [...] means taking the set average and Ω_{0, l}And C_lIs expressed as follows.
[0023]
[Equation 5]

However, in the derivation of equations (4.1) and (4.3), noise and signal uncorrelation and uncorrelation between signals of different users were used. In the equation (4.1), I is a unit matrix, σ²Is the variance of the noise.
[0024]
Here, E [C] = {E [c_{i, j}]} Is actually as follows.
[0025]
[Formula 6]

Substituting the above equations (5.1) and (5.2) into equation (4.1) gives the following equation.
[0026]
[Expression 7]

Here, the equation (6.1) is modified as follows.
[0027]
[Equation 8]

The above correlation matrix R_kEigenvalue decomposition, eigenvalue λ_iIs as follows.
[0028]
[Equation 9]

Equations (8.1) and (8.2) mean that equation (7.1) can be modified as follows.
[0029]
[Expression 10]

From the equations (8.1) and (8.2), λ₁> Λ₂Therefore, the eigenvector corresponding to the maximum value of the eigenvalues of the correlation matrix is a carrier vector or a frequency offset vector. Therefore, the frequency can be estimated by measuring the frequency of the eigenvector corresponding to the maximum eigenvalue of the correlation matrix. Also, as can be seen from the equations (8.1) and (8.2), the eigenvalue λ including the coherent component of the desired signal is obtained by using the eigenvalue decomposition.₁And the eigenvalue λ without coherent components₂It can be seen that the frequency can be estimated with high accuracy.
[0030]
Based on the above principle consideration, the present invention is characterized by the following points.
[0031]
The invention of claim 1 is a frequency offset estimation method in a broadcast or communication reception method for estimating a carrier frequency of a received signal, converting from a carrier band to a baseband based on an estimation result, and demodulating the signal, Multiplication processing for multiplying the received signal by a known signal, delay processing for outputting the output of the multiplication processing at predetermined time intervals, and correlation matrix calculation for calculating a vector correlation matrix using the output of the delay processing as an element And eigenvalue decomposition processing for obtaining an eigenvalue of a correlation matrix that is an output result of the correlation matrix calculation process, and outputting an eigenvector corresponding to the maximum eigenvalue as a waveform of a carrier wave.
[0032]
The invention of claim 2 is a frequency offset estimation method in a broadcast or communication reception method for estimating a carrier frequency of a received signal, converting from a carrier band to a baseband based on an estimation result, and demodulating the signal, Multiplication processing for multiplying the received signal by a known signal, delay processing for outputting the output of the multiplication processing at predetermined time intervals, and correlation matrix calculation for calculating a vector correlation matrix using the output of the delay processing as an element And eigenvector frequency estimation processing for obtaining an eigenvalue of a correlation matrix, which is the result output of the correlation matrix calculation process, and estimating and outputting the maximum value of the frequency component included in the eigenvector corresponding to the maximum eigenvalue. is there.
[0033]
The invention of claim 3 is a frequency offset estimator in a broadcast or communication receiving facility that estimates a carrier frequency of a received signal, converts the carrier wave band to a baseband band based on the estimation result, and demodulates the signal, Multiplication means for multiplying the received signal by a known signal, delay means for outputting the output of the multiplication means at predetermined time intervals, and correlation matrix calculation means for calculating a correlation matrix of a vector having the output of the delay means as elements. And eigenvalue decomposing means for obtaining eigenvalues of the correlation matrix, which is an output of the correlation matrix calculating means, and outputting an eigenvector corresponding to the maximum eigenvalue as a waveform of the carrier wave.
[0034]
The invention of claim 4 is a frequency offset estimator in a broadcast or communication receiving facility for estimating a carrier frequency of a received signal, converting from a carrier band to a baseband based on an estimation result, and demodulating the signal, Multiplication means for multiplying the received signal by a known signal, delay means for outputting the output of the multiplication means at predetermined time intervals, and correlation matrix calculation means for calculating a correlation matrix of a vector having the output of the delay means as elements. And eigenvector frequency estimation means for obtaining an eigenvalue of a correlation matrix, which is an output of the correlation matrix calculation means, and estimating and outputting a maximum value of a frequency component included in an eigenvector corresponding to the maximum eigenvalue. is there.
[0035]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0036]
<First Embodiment>
FIG. 1 shows the configuration of a frequency offset estimator according to the first embodiment of the present invention. In the figure, 100 is an input terminal, 101 is a multiplier, 102 is a pilot signal storage element (pilot symbols), 103 is a delay circuit, 104 is an eigenvalue decomposition unit, 105 is a correlation matrix calculation unit, and 106 is an output terminal for an estimated frequency. It is. In the figure, the input signal at the input terminal 100 is multiplied by a known signal output from the pilot signal storage element 102, and y in equation (2) is obtained._kGet. This y_kAre input to the delay circuit 103, the signal groups y having different delay times from the delay circuit 103, respectively._k, y_k-1, ..., y_{k-L + 1}Is output. This group of signals is a vector Y_kIs equal to the element of
[0037]
Accordingly, the correlation matrix calculation unit 105 performs this vector Y_kIs used to calculate the correlation matrix R_kGet.
[0038]
## EQU11 ##

In equation (4.1), the correlation matrix is Y_kY_k ^HHowever, in actuality, it is obtained as a time average based on Ergod's theorem. This correlation matrix is a Hermitian matrix as can be seen from the equation (11). Therefore, the eigenvalue decomposition can be obtained by various methods such as the Jacobi method and the QR method. The eigenvalue decomposition unit 104 performs this eigenvalue decomposition process. (Kurose; “Subroutine library for Fortran90: numerical calculation, statistical calculation, 3D CG library”, Morikita Publishing, 1998).
[0039]
The eigenvalue decomposition unit 104 extracts only the eigenvector corresponding to the maximum eigenvalue from the eigenvalues and eigenvectors thus obtained, and outputs them from the output terminal 106 as an estimated value of the frequency offset.
[0040]
A detailed configuration of the delay circuit 103 in the frequency offset estimator is shown in FIG. In the figure, 201 is an input terminal, 202 to 204 are delay elements, and 205 to 208 are output terminals of a signal group. Input terminal 201 to y_kIs input, it is delayed and y_k, y_k-1, ..., y_{k-L + 1}Are output from output terminals 205-208.
[0041]
<Second Embodiment>
Next, FIG. 3 shows a frequency offset estimator according to the second embodiment of the present invention. In this figure, 111 is an input terminal, 112 is a multiplier, 113 is a delay circuit, 114 is a pilot signal storage element, 115 is a correlation matrix calculation unit, 116 is an eigenvalue decomposition unit, 117 is an eigenvector frequency estimation unit, and 118 is an estimation This is the frequency output terminal.
[0042]
In the second embodiment, the eigenvalue is obtained from the correlation matrix of the received signal vector as in the first embodiment shown in FIG. 1, and a vector corresponding to the maximum eigenvalue is generated. Since the generated vector is not necessarily a sine wave, the highest frequency component of the eigenvector is obtained by the frequency estimator 117 of the eigenvector and output from the output terminal 118 as an estimated value of the frequency offset.
[0043]
FIG. 4 shows a detailed configuration of the eigenvector frequency estimation unit 117 described above. In the same figure, 221-1 to 221-4 are input terminals of each element of the eigenvector from the eigenvalue decomposition unit 116, 222-1 to 222-4 are buffer circuits for storing the eigenvector, 223 to 229 are multipliers, 230 is a terminal for inputting a fixed value “1”, 231 is an adder, 234 is an adaptive control unit, and 235 is an output terminal. The adaptive control unit 234 also uses the eigenvector v₁= [V_1,0, v_1,1, ..., v_{1, L-1}], The following calculation is repeated.
[0044]
[Expression 12]

However, μ in the expression (12.3) represents a minute coefficient called a step size parameter, and Im [...] Represents a function for extracting only the imaginary part.
[0045]
After repeating the above calculation a predetermined number of times m,
[Formula 13]

Is output from the terminal 235 as an estimated frequency.
[0046]
<Third Embodiment>
A frequency offset estimator according to the third embodiment of the present invention is shown in FIG. In general, in a radio system, an approximate frequency is known because a signal is arranged in a known radio frequency band. However, depending on the case, there may be a large deviation from the assumed initial frequency due to incompleteness of the device. The frequency offset estimation circuit according to the third embodiment is an effective circuit when the estimated frequency is greatly deviated from the estimated initial value.
[0047]
In the figure, 121 is an input terminal, 122 is a multiplier, 123 is a pilot signal storage element, 124 is a delay circuit, 125 is a correlation matrix calculation unit, 126 is an eigenvalue decomposition unit, 127 is a frequency estimation unit for eigenvectors, and 128 is a DFT. A circuit 129 is an output terminal for an estimated value of the frequency offset, and 130 is a maximum value detection unit.
[0048]
The circuit of the third embodiment performs the same processing as that of the circuit of the second embodiment shown in FIG. 3, except that the eigenvector corresponding to the maximum eigenvalue is input to the DFT circuit 128 and is different from the DFT circuit 128. Processing, the maximum value detector 130 determines the frequency component corresponding to the maximum value as an initial value in frequency estimation of the eigenvector,
[Expression 14]

The eigenvector frequency estimation unit 127 performs frequency estimation using this.
[0049]
In principle, the frequency estimation of the eigenvector by the frequency estimation unit 117 shown in FIG. 4 employed in the second embodiment is highly accurate, but is quite different when the initial value is too large and is different from the actual frequency. May be output. Therefore, in the third embodiment, this frequency can be overcome by using the frequency roughly estimated by the DFT circuit 128 as an initial value.
[0050]
<Application example 1>
FIG. 6 shows a receiver using the frequency offset estimator of the present invention. In this figure, 131 is a carrier band signal input terminal, 132 is a local oscillator, 133 to 135 are multipliers, 136 is a delay element, 137 is the frequency offset estimator of the second embodiment shown in FIG. 3, or FIG. The frequency offset estimator 138 of the third embodiment shown in FIG. 1 is a signal output terminal 138, and is a band pass filter.
[0051]
In this receiver, the received signal is once frequency-converted to near the baseband, and the frequency offset that is the error between the local oscillator that is the IF frequency and the carrier frequency of the received signal is estimated. That is, the received signal input from the input terminal 131 is multiplied by the signal from the local oscillator 132 by the multiplier 133 to obtain a signal in the IF frequency band. This signal is branched after passing through the band-pass filter 139, one of which is input to the frequency offset estimator 137 to estimate the frequency offset. Then, the multiplier 135 and the delay element 136 perform phase integration processing on the estimated frequency offset, and the phase error due to the frequency offset is output from the multiplier 135, and the output signal of the branched bandpass filter 139 and By multiplying by the multiplier 134, a complete baseband signal is obtained and output from the signal output terminal 138.
[0052]
<Application example 2>
FIG. 7 shows a receiving facility of a CDMA (Code Division Multiple Access) base station employing the frequency offset estimator of the present invention. In the figure, 140 is a band pass filter, 141 is a received signal input terminal, 142 is a local oscillator, 143 to 145 are multipliers, 146 is a delay element, 147 is a frequency offset estimator, 148 is an external pilot signal storage circuit, 149 ˜151 are frequency offset compensators (AFC), 152˜154 are CDMA receivers, and 155˜157 are demodulated signal output terminals. In CDMA, several users communicate using the same radio band at the same time. Therefore, several users' signals are mixed in the received signal. Strictly speaking, the radio carrier frequency is slightly different for each user. The receiving facility in FIG. 7 is configured to demodulate all these user signals.
[0053]
In the case of the second application example, as in the first application example shown in FIG. 6, the received signal converted to the IF frequency band by the local oscillator 142 is subjected to the frequency offset compensation of # 1 to #N. Input to the devices 149-151.
[0054]
Different pilot signals are downloaded from the external pilot signal storage circuit 148 to the pilot signal storage elements in each of the frequency offset compensators 149 to 151. Therefore, in each of the frequency offset compensators 149 to 151, the frequency offset estimator 147, the multipliers 144 and 145, and the delay element 146 remove the IF frequency of the received signal as described with reference to FIG. It should be noted that in the frequency offset compensators 149 to 151, only a signal including the pattern downloaded by the pilot signal storage circuit 148, that is, a desired signal is converted to baseband, and other signals are converted to baseband. There is no guarantee that it will be converted to. The outputs of the frequency offset compensators 149 to 151 are input to the receivers 152 to 154, respectively, to obtain demodulated signals.
[0055]
FIG. 8 shows the configuration of CDMA receivers 152 to 154 used in the CDMA reception facility of FIG. In the figure, 161 is an input terminal, 162 is a delay circuit, 163 to 165 are matched filters for despreading, 166 to 168 are multipliers, 169 is an adder, 170 is an adaptive control unit, and 171 is a demodulated signal output terminal. is there. The configuration of this receiver is a so-called Rake receiver, and it is possible to maximize the SNR (signal-to-noise ratio) by adding in-phase addition of the delayed wave in addition to the preceding wave.
[0056]
As the configuration of the adaptive control unit 170, since the frequency offset has already been removed, an LMS (Least Mean Square) method, an RLS (Recursive Least Squares) method, and the like, which are algorithms based on the linear estimation theory, can be applied. The following shows an algorithm when the LMS method is applied as an example. First, each output of the matched filters 163 to 165 is set to Z_k= [Z_{k, 1} z_{k, 2} ... z_{k, M}]^TAnd the coefficient vector from the adaptive control unit 170 is W_k= [W_{k, 1} w_{k, 2} … W_{k, M}]^TAnd Where M is the number of matched filters. At that time,
[Expression 15]

The demodulated signal can be obtained by the algorithm. However, D_kIs W_k ^HZ_kThis is a known symbol pattern called a discriminant value or training signal.
[0057]
FIG. 9 shows a specific circuit configuration of the matched filters 163 to 165 shown in FIG. The figure shows an example of the configuration when the spreading factor is 4. In the figure, 240 is an input terminal, 241 to 243 are delay elements, 244 to 247 are spread code input terminals, 248 to 251 are multipliers, 252 is an adder, and 253 is an output terminal.
[0058]
<Application example 3>
FIG. 10 shows the configuration of a multiuser reception facility to which the frequency offset estimator of the present invention is applied. In this figure, 181 is a received signal input terminal, 182 is a local oscillator, 183 is a multiplier, 184 is a band pass filter, 185 is a frequency offset estimator, 186 is an external pilot signal storage circuit, and 187 to 190 are pilot signal download functions. A frequency offset estimator 191 is a multi-user receiver, and 192 to 194 are demodulated signal output terminals.
[0059]
In the multi-user reception facility having this configuration, the frequency offset of the signal included in the received signal is estimated by the frequency offset estimators 187 to 190 with a pulse signal download function. The estimated value of the frequency offset for each user is input to the multiuser receiver 191. The multi-user receiver 191 collectively demodulates each user's signal using an estimated value of each user's frequency offset and a known signal embedded between information sequences called training sequences. The demodulated signal is output from output terminals 192 to 194.
[0060]
FIG. 11 shows a first configuration example of the multiuser receiver 191 shown in FIG. The figure shows an example of the configuration of a 4-user receiver. In the figure, 261 is an input terminal, 262 to 265 are input terminals of estimated offset frequencies, 266 is a subtractor, 267 is a squaring circuit, 268 is a maximum likelihood estimator (MLE), 269 to 272 are A replica generator 273 is an adaptive control unit that estimates a transmission line impulse response of each user, and 274 to 277 are demodulation signal output terminals. In the figure, a replica is generated by adding frequency offset information to an ordinary multiuser receiver, and a demodulated signal with the highest likelihood is obtained.
[0061]
FIG. 12 shows a specific configuration of the replica generator employed in the multiuser receiver of FIG. In the figure, 281 is an input terminal, 282 is a delay circuit, 283 to 288 are multipliers, 289 is an impulse response input terminal from the adaptive control unit 273 in FIG. 11, 290 is an adder, 291 is a delay element, and 292 is an offset. A frequency input terminal 293 is a replica output terminal.
[0062]
In this replica generator, the provisional determination value from the input terminal 281 is convolved by a FIR (Finite Impulse Response) filter composed of a delay circuit 282 and multipliers 283 to 286 to obtain a baseband replica. On the other hand, by integrating the offset frequency signal from the offset frequency input terminal 292 by the multiplier 288 and the delay element 291, the phase fluctuation component due to the offset frequency is calculated. Then, a multiplier 287 multiplies the phase fluctuation component by a baseband replica to generate a received signal replica, and outputs the received signal replica from the output terminal 293 to the MLE circuit 268 of FIG.
[0063]
FIG. 13 shows a configuration example of the MLE circuit 268 applied to the multiuser receiver 191 shown in FIG. The figure shows a maximum likelihood estimation circuit for four users using a BPSK modulated signal. In this figure, 301 is a signal input terminal from the squaring circuit 267 of FIG. 11, 302 is a reset signal input terminal, 303 to 307 are switch circuits, 308 is a subtractor, 309 is a discriminator, 310 is a delay element, 311 is binary A counter, 312 is a clock input terminal, 313, 314, 319, and 320 are demodulated signal output terminals, and 315 to 318 are provisional judgment value output terminals.
[0064]
In this MLE circuit 268, the binary counter 311 is driven by a clock to generate all the signals of four users subjected to BPSK modulation. A 4-user replica is generated based on the generated temporary determination value, and the square Euclidean distance between the replica and the received signal is input to the switch circuit 304. A minimum value detector composed of a subtracter 308, a delay element 310, and a switch circuit 303 obtains a provisional determination value having the minimum square Euclidean distance, which is selected by the switch circuit 303 and output as a demodulated signal.
[0065]
Similarly to the CDMA receivers 152 to 154 in FIG. 7, the adaptive control unit 273 in the multiuser receiver 191 shown in FIG. 11 also has an algorithm based on linear estimation theory such as LMS and RLS since the frequency offset of each user is known. Is applicable.
[0066]
FIG. 14 shows a second configuration example of the multiuser receiver 191 shown in FIG. This is provided with a multi-beam antenna which receives a plurality of antenna elements, K-branches them, and adaptively forms a beam at each of the K branch destinations. In this figure, 330 is an input terminal for an offset frequency estimated by the frequency offset estimator in FIG. 10, 331 is an input terminal for received signals received by N antenna elements, and 332 to 334 are beamformers (BFN: Beam Forming Network), 335 to 337 are demodulation signal output terminals.
[0067]
In the multi-user receiver having this configuration, the i-th beamformer cancels the signals of the first to i-1th users, and thereafter, the SINR (Signal to Interference and Noise Ratio) of the output signal becomes maximum. Control the directivity of the antenna. At this time, in order to cancel the signal accurately, by applying phase rotation to each demodulated signal based on the estimated frequency offset, an algorithm based on linear estimation theory can be used for its control algorithm, and high interference suppression characteristics Can be achieved. In the actual communication path, the frequency offset differs for each user. Therefore, by performing interference cancellation using the configuration shown in FIG. 14, high communication quality can be achieved even in an actual communication path.
[0068]
FIG. 15 shows a configuration example of the j-th beam former BNF (j) in the multiuser receiver 191 of FIG. In the figure, 341 to 344 are signal input terminals from each antenna element, 345 to 351, 360, 362 and 367 to 369 are multipliers, 352 is an adder, 353 is a subtractor, 354 is a determiner, and 355 is a switch. 356 is a known signal input terminal called a training signal, 357 is a demodulated signal output terminal, 358 is an adaptive control unit (Adaptive Weight Control), 371 to 373 are signals demodulated by the j-1st BFN (j-1) , 364 is an input terminal of the estimated frequency offset, 361 is a delay element, 363 is an integrator, 365 is a complex conjugate (·)^*Represents.
[0069]
The adaptive control unit 358 uses the output signal from the demodulated signal output terminal 357 as a desired signal, the output signal from the subtractor 353 as an error signal, the outputs from the multipliers 367 to 369 and the signals from the input terminals 341 to 344 as input signals. The coefficients of the multipliers 345 to 351 are adaptively estimated using a linear estimation algorithm such as an LMS or RLS algorithm.
[0070]
FIG. 16 shows a frequency offset estimation error characteristic by the RLS adaptive phase control method which is a conventional technique. The mean square of the estimation error is the vertical axis, and the horizontal axis represents the number of received symbols. The figure shows the characteristics when each channel is a two-path model (CCI + ISI) in addition to the AWGN channel and two-path model (ISI), one-wave co-channel interference (CCI), and one-wave co-channel interference. In the figure, CNR = 0 dB, frequency offset ΔfT = 10^-3The characteristics were obtained under the following conditions. However, T represents a symbol period. As can be seen from this characteristic, in the RLS adaptive phase control method, high-accuracy estimation is possible in the AWGN channel, but degradation starts as soon as interference enters, and 10 C is obtained in the CCI + ISI transmission path.^-2Only a certain degree of estimation accuracy is obtained.
[0071]
FIG. 17 shows an example of characteristics of the present invention. The conditions were the same as those in the case of characteristic acquisition in the conventional example shown in FIG. Characteristics were evaluated using only 140 symbols of training signals. Frequency offset ΔfT = 10^-3In the case of, in the 140 symbol training signal section, the phase does not rotate by 2π, so a slight estimation error is seen, but the estimation error is 3 × 10^-FourIt can be seen that this is an excellent characteristic to drive up to.
[0072]
【The invention's effect】
According to the present invention, frequency offset estimation can be performed with high accuracy in a severe co-channel interference / intersymbol interference environment. By using this high-accuracy frequency offset information, not only the communication quality of the base station receiver in the mobile communication system is improved, but also an adaptive array having a strong interference compensation function, a multi-user receiver, etc. are used for actual communication. Make it applicable.
[0073]
That is, in order to use a multi-user receiver that requires channel estimation based on linear estimation theory for actual communication, a frequency offset estimator that accurately estimates a different frequency offset for each user even under poor environmental conditions. However, by solving this problem, the present invention makes it possible to introduce a multi-user receiver that enables a significant increase in frequency utilization efficiency into a commercial system, thereby improving frequency utilization efficiency. Can contribute.
[Brief description of the drawings]
FIG. 1 is a block diagram of a frequency offset estimator according to a first embodiment of this invention.
FIG. 2 is a block diagram of a delay circuit in the above embodiment.
FIG. 3 is a block diagram of a frequency offset estimator according to a second embodiment of this invention.
FIG. 4 is a block diagram of a frequency estimation unit for eigenvectors in the above embodiment.
FIG. 5 is a block diagram of a frequency offset estimator according to a third embodiment of this invention.
FIG. 6 is a block diagram of a reception facility according to the fourth embodiment of this invention.
FIG. 7 is a block diagram of a reception facility according to a fifth embodiment of this invention.
FIG. 8 is a block diagram of a receiver in the fourth and fifth embodiments.
FIG. 9 is a block diagram of a matched filter in the receiver.
FIG. 10 is a block diagram of a reception facility according to a sixth embodiment of this invention.
FIG. 11 is a block diagram of a multiuser receiver in the sixth embodiment.
FIG. 12 is a block diagram of a replica generator in the multiuser receiver.
FIG. 13 is a block diagram of a maximum likelihood estimation circuit (MLE) in the multiuser receiver.
FIG. 14 is a block diagram showing another configuration example of the multiuser receiver in the sixth embodiment.
FIG. 15 is a block diagram of a j-th beamformer BFN (j) in the multiuser receiver.
FIG. 16 is a graph showing characteristics of a conventional example.
FIG. 17 is a graph showing characteristics of the present invention.
FIG. 18 is a block diagram of a conventional Costas type carrier regenerator.
FIG. 19 is a block diagram of a conventional quadrature detector.
FIG. 20 is a block diagram of a conventional inverse transform carrier regenerator.
FIG. 21 is a block diagram of a conventional delay detector with a frequency offset compensation function.
FIG. 22 is a block diagram of a conventional RLS phase control circuit.
FIG. 23 is a block diagram of a frequency offset estimator to which a conventional DFT is applied.
[Explanation of symbols]
102 Pilot signal storage element
103 Delay circuit
104 Eigenvalue decomposition unit
105 Correlation matrix calculator
114 Pilot signal storage element
113 Delay circuit
115 Correlation matrix calculator
116 Eigenvalue decomposition unit
117 Eigenvector Frequency Estimator
123 Pilot signal storage element
124 delay circuit
125 correlation matrix calculator
126 Eigenvalue decomposition unit
127 Eigenvector Frequency Estimator
128 DFT
130 Maximum value detector

Claims (4)

A frequency offset estimation method in a broadcast or communication reception method for estimating a carrier frequency of a received signal, converting from a carrier band to a baseband based on an estimation result, and demodulating the signal,
A multiplication process for multiplying the received signal by a known signal;
A delay process for outputting the output of the multiplication process at predetermined time intervals;
A correlation matrix calculation process for calculating a correlation matrix of a vector having the output of the delay process as an element;
A frequency offset estimation method comprising: eigenvalue decomposition processing for obtaining an eigenvalue of a correlation matrix, which is an output resulting from the correlation matrix calculation processing, and outputting an eigenvector corresponding to the maximum eigenvalue as a carrier waveform. A frequency offset estimation method in a broadcast or communication reception method for estimating a carrier frequency of a received signal, converting from a carrier band to a baseband based on an estimation result, and demodulating the signal,
A multiplication process for multiplying the received signal by a known signal;
A delay process for outputting the output of the multiplication process at predetermined time intervals;
A correlation matrix calculation process for calculating a correlation matrix of a vector having the output of the delay process as an element;
A frequency offset estimation method comprising: an eigenvector frequency estimation process that obtains an eigenvalue of a correlation matrix that is output as a result of the correlation matrix calculation process, and estimates and outputs a maximum value of a frequency component included in an eigenvector corresponding to the maximum eigenvalue. A frequency offset estimator in a broadcast or communication receiving facility that estimates a carrier frequency of a received signal, converts the carrier band from a baseband to a baseband based on an estimation result, and demodulates the signal,
Multiplying means for multiplying the received signal by a known signal;
Delay means for outputting the output of the multiplication means at predetermined time intervals;
Correlation matrix calculation means for calculating a correlation matrix of a vector having the output of the delay means as an element;
A frequency offset estimator comprising: eigenvalue decomposition means for obtaining an eigenvalue of a correlation matrix, which is an output of the correlation matrix calculation means, and outputting an eigenvector corresponding to the maximum eigenvalue as a carrier waveform. A frequency offset estimator in a broadcast or communication receiving facility that estimates a carrier frequency of a received signal, converts the carrier band from a baseband to a baseband based on an estimation result, and demodulates the signal,
Multiplying means for multiplying the received signal by a known signal;
Delay means for outputting the output of the multiplication means at predetermined time intervals;
Correlation matrix calculation means for calculating a correlation matrix of a vector having the output of the delay means as an element;
A frequency offset estimator comprising: an eigenvector frequency estimating unit that obtains an eigenvalue of a correlation matrix that is an output of the correlation matrix calculating unit and estimates and outputs a maximum value of a frequency component included in an eigenvector corresponding to the maximum eigenvalue .

Images