JP4438280B2

JP4438280B2 - Transcoder and code conversion method

Info

Publication number: JP4438280B2
Application number: JP2002317204A
Authority: JP
Inventors: 一範小澤
Original assignee: NEC Corp
Current assignee: NEC Corp
Priority date: 2002-10-31
Filing date: 2002-10-31
Publication date: 2010-03-24
Anticipated expiration: 2022-10-31
Also published as: AU2003271119A1; EP1564723A1; HK1077913A1; EP1564723A4; EP1564723B1; CN1708786A; CA2504174A1; WO2004040552A1; JP2004151424A; DE60321712D1; KR100715014B1; CN100498933C; KR20050061579A

Description

【０００１】
【発明の属する技術分野】
本発明は、第１の符号化方式による符号と第２の符号化方式による符号とを相互変換するトランスコーダに関し、第１の符号から第２の符号に変換する際に、信号の周波数帯域を拡張化した上で変換するトランスコーダに関するものである。
【０００２】
【従来の技術】
低ビットレートで符号化して再生した音声信号の周波数帯域を、帯域拡張のための補助情報を送信側から伝送することなく、受信側で拡張させる方式が知られている（例えば非特許文献１）。
【０００３】
【非特許文献１】
P.Jax, P.Vary，"Wideband extension of telephone speech using hidden markov model," Proc. IEEE Speech Coding Workshop, pp.133-135,2000年
【０００４】
この従来方式では、受信側でHMM(Hidden Markov Model)を用いて帯域拡張した後のフィルタ係数を探索している。
【０００５】
一方、第１の符号化方式による符号と第２の符号化方式による符号とを相互変換するトランスコーダにおいて、第１の符号から第２の符号に変換する際に、信号の帯域を拡張化した上で変換するトランスコーダに関しては、前例がない。
【０００６】
【発明が解決しようとする課題】
P.Jax, P.Varyによる前記文献（上記非特許文献１）に記載された従来法では、広帯域音声のスペクトル包絡やフィルタ係数のHMMによるモデル化が必要であり、次のような問題点があった。
【０００７】
すなわち、あらかじめオフラインで多量の音声データベースからHMMモデルのパラメータを決定しておく必要があり、これに多大な計算時間、コストが必要である。
【０００８】
また、受信側でリアルタイムに帯域拡張処理を行う際に、HMMモデルによる探索が必要でこれに多くの演算量が必要であった。
【０００９】
したがって、本発明の主たる目的は、第１の符号化方式による符号と第２の符号化方式による符号とを相互変換するトランスコーダにおいて、第１の符号から第２の符号に変換する際に、信号の帯域を拡張化した上で変換する際に、比較的少ない演算量で、音質の良好な帯域拡張化変換を施すことを可能とするトランスコーダ及び符号変換方法を提供することにある。
【００１０】
【課題を解決するための手段】
前記目的を達成する本発明は、第１の符号化方式による符号と第２の符号化方式による符号とを相互変換するトランスコーダであって、第１のアスペクトによれば、第１の符号化方式により符号化された符号を入力して第１の符号化方式により復号化しスペクトル特性を表すスペクトルパラメータを計算するスペクトルパラメータ計算部と、雑音信号を発生する雑音発生部と、前記スペクトルパラメータの周波数をシフトさせた上でフィルタ係数を求める係数計算部と、前記雑音発生部の出力に適切なゲインを与えるゲイン部と、前記ゲイン部の出力を前記係数を用いた構成した合成フィルタに通し帯域拡張信号を再生する合成フィルタ部と、前記入力信号の標本化周波数を変換した上で前記合成フィルタ部の出力信号を加算して出力する加算部と、前記加算部の出力信号を第２の符号化方式により符号化して第２の符号を出力する構成とされる。
【００１１】
また、本発明に係るトランスコーダは、第２のアスペクトによれば、第１の符号化方式による符号と第２の符号化方式による符号とを相互変換するトランスコーダにおいて、第１の符号化方式により符号化された符号を入力して第１の符号化方式により復号化しスペクトル特性を表すスペクトルパラメータを計算するスペクトルパラメータ計算部と、前記入力信号からピッチ周期を計算し前記ピッチ周期と過去の音源信号に基づき適応コードブック成分を発生させる適応コードブック部と、雑音信号を発生する雑音発生部と、前記スペクトルパラメータの周波数をシフトさせた上でフィルタ係数を求める係数計算部と、前記雑音発生部の出力と前記適応コードブック部の出力の少なくとも一方に適切なゲインを与えた上で加算し音源信号を出力するゲイン部と、前記係数を用いて構成した合成フィルタに前記音源信号を入力し帯域拡張信号を再生する合成フィルタ部と、前記再生信号の標本化周波数を変換した上で前記合成フィルタ部の出力信号を加算して出力する加算部と、前記加算部の出力信号を第２の符号化方式により符号化して第２の符号を出力する構成とされる。
【００１２】
また、本発明に係るトランスコーダは、第３のアスペクトによれば、第１の符号化方式による符号と第２の符号化方式による符号とを相互変換するトランスコーダにおいて、第１の符号化方式により符号化された符号を入力して第１の符号化方式により復号化しスペクトル特性を表すスペクトルパラメータを計算するスペクトルパラメータ計算部と、前記入力信号からピッチ周期を計算し前記ピッチ周期と過去の音源信号に基づき適応コードブック成分を発生させる適応コードブック部と、雑音信号を発生する雑音発生部と、前記スペクトルパラメータの周波数をシフトさせた上でフィルタ係数を求める係数計算部と、前記雑音発生部の出力と前記適応コードブック部の出力の少なくとも一方に適切なゲインを与えた上で加算し音源信号を出力するゲイン部と、前記ピッチ周期を用いて前記音源信号に対しピッチプリフィルタを通し、前記係数を用いて構成した合成フィルタに前記ピッチプリフィルタ出力信号を入力し帯域拡張信号を再生する合成フィルタ部と、前記再生信号の標本化周波数を変換した上で前記合成フィルタ部の出力信号を加算して出力する加算部と、前記加算部の出力信号を第２の符号化方式により符号化して第２の符号を出力する構成とされる。
【００１３】
また、本発明によれば、適応コードブック部の出力をあらかじめ定められたカットオフ周波数の低域通過型フィルタに通す構成としてもよい。
【００１４】
さらに、本発明によれば、前記係数に重み付けを施した重み付け係数を用いてポストフィルタを構成し、前記合成フィルタ部の出力信号を前記ポストフィルタに通して帯域拡張信号を再生する構成としてもよい。
【００１５】
本発明に係る方法の一つのアスペクトによれば、第１の符号化方式による符号と第２の符号化方式による符号とを相互に変換するトランスコーダの符号変換方法において、第１の符号化方式により符号化された符号を第１の復号化方式により復号化し復号信号を出力するステップと、
前記復号信号からスペクトル特性を表すスペクトルパラメータを計算して出力するステップと、
前記スペクトルパラメータの周波数をシフトさせた上でフィルタ係数を求めて出力するステップと、
雑音発生部からの出力信号にゲインを与えるステップと、
前記フィルタ係数を用いて構成される合成フィルタに、前記ゲインを与えた出力信号を通し、帯域変換に必要な帯域の信号を出力するステップと、
前記復号信号をあらかじめ定められた標本化周波数で変換した信号と前記合成フィルタの出力信号とを加算するステップと、
前記加算結果を前記第２の符号化方式により符号化し第２の符号を得て出力するステップと、を含む。
【００１６】
あるいは、本発明に係る方法の他のアスペクトによれば、前記第１の符号化方式により符号化された符号を第１の復号化方式により復号化し復号信号を出力するステップと、
前記復号信号からスペクトル特性を表すスペクトルパラメータを計算して出力するステップと、
前記復号信号からピッチ周期を計算し前記ピッチ周期と過去の音源信号に基づき適応コードブック成分を発生させるステップと、
前記スペクトルパラメータの周波数をシフトさせた上でフィルタ係数を求めて出力するステップと、
雑音発生部の出力と前記適応コードブック成分の少なくとも一方にゲインを与えた上で加算し音源信号を出力するステップと、
前記フィルタ係数を用いて構成される合成フィルタに、前記音源信号を通し帯域変換に必要な帯域の信号を出力するステップと、
前記復号信号をあらかじめ定められた標本化周波数で変換した信号と前記合成フィルタの出力信号とを加算するステップと、
前記加算結果を前記第２の符号化方式により符号化し第２の符号を得て出力するステップと、を含む。
【００１７】
あるいは、本発明に係る方法の他のアスペクトによれば、前記第１の符号化方式により符号化された符号を第１の復号化方式により復号化し復号信号を出力するステップと、
前記復号信号からスペクトル特性を表すスペクトルパラメータを計算して出力するステップと、
前記復号信号からピッチ周期を計算し前記ピッチ周期と過去の音源信号に基づき適応コードブック成分を発生させるステップと、
前記スペクトルパラメータの周波数をシフトさせた上でフィルタ係数を求めて出力するステップと、
雑音発生部の出力と前記適応コードブック成分の少なくとも一方にゲインを与えた上で加算し音源信号を出力するステップと、
前記ピッチ周期を用いて前記音源信号に対しピッチプリフィルタを行うステップと、
前記フィルタ係数を用いて構成される合成フィルタに、前記ピッチプリフィルタ処理された信号を通し帯域変換に必要な帯域の信号を出力するステップと、
前記復号信号をあらかじめ定められた標本化周波数で変換した信号と前記合成フィルタの出力信号とを加算するステップと、
前記加算結果を前記第２の符号化方式により符号化し第２の符号を得て出力するステップと、を含む。
【００１８】
本発明に係る方法の他のアスペクトによれば、ピッチプリフィルタで前記ピッチ周期を用いて、前記ゲイン部からの前記音源信号に対しプリフィルタ処理を行い、合成フィルタに、ピッチプリフィルタからの出力信号を入力するステップを有する構成としてもよい。
【００１９】
本発明に係る方法の他のアスペクトによれば、前記計数計算部からのフィルタ係数に重み付けを施した重み付け係数を用いて構成されたポストフィルタに、前記合成フィルタ部の出力信号を入力するステップを有する構成としてもよい。
【００２０】
さらに本発明に係る方法の他のアスペクトによれば、適応コードブック部からの出力信号のかわりに、ピッチ周期を用いて周期信号を発生させる周期信号発生部の出力をゲイン部に供給するようにしてもよい。
【００２１】
【発明の実施の形態】
本発明の実施の形態について図面を参照して説明する。以下では、第１の符号は４ｋＨｚ帯域の狭帯域入力信号に対して符号化されたものであり、トランスコーダにおいて、これを５ｋＨｚ帯域あるいは７ｋＨｚ帯域の信号に帯域拡張化した上で、第２の符号化方式により符号化して第２の符号を得ることを想定する。
【００２２】
図１は、本発明に係るトランスコーダの第１の実施の形態の構成を示すブロック図である。図１を参照すると、トランスコーダは、第１の復号回路１０５と、スペクトルパラメータ計算回路１００と、雑音発生回路１２０と、係数計算回路１３０と、合成フィルタ回路１７０と、標本化周波数変換回路１８０と、加算器１９０と、第２の符号化回路１９５と、有声/無声判別回路２００と、ゲイン調整回路３１０と、ゲイン回路１４０とを備えて構成されている。
【００２３】
第１の復号回路１０５は、第１の符号化方式による符号を入力し、第１の復号化方式により復号化して、復号信号x(n)を出力する。
【００２４】
スペクトルパラメータ計算回路１００は、復号信号x(n)をフレームに分割（例えば１０ｍｓ）し、フレーム毎に、あらかじめ定められた次数Pのスペクトルパラメータを計算する。ここで、スペクトルパラメータは、フレーム毎の音声信号のスペクトル概形を表すパラメータであり、この計算には、周知のＬＰＣ（線形予測符号化）分析等を用いることができる。さらに、スペクトルパラメータ計算部１００では、ＬＰＣ分析により計算された線形予測係数αi(i=1,…P)を、量子化や補間に適したＬＳＰ（Line Spectrum Pair：線スペクトル対）パラメータに変換し出力する。ここで、線形予測係数からＬＳＰへの変換は、例えば次の論文が参照される（非特許文献２）。
【００２５】
【非特許文献２】
菅村、板倉：”線スペクトル対（LSP）音声分析合成方式による音声情報圧縮”、電子通信学会論文誌、J64-A、pp.599-606、1981年
【００２６】
係数計算回路１３０は、スペクトルパラメータ計算回路１００から出力されたスペクトルパラメータを入力し、帯域拡張された信号の係数に変換する。この変換には、例えば、ＬＳＰの周波数を単純に高い周波数へシフトさせる手法、非線形変換手法、線形変換手法などの周知の方法を用いることができる。ここでは、ＬＳＰパラメータの全部または一部を使用して、ＬＳＰの存在周波数帯域を高い周波数域にシフトさせた上で、次数Ｐの線形予測係数に変換し、合成フィルタ回路１７０に出力する。
【００２７】
雑音発生部１２０は、平均振幅があらかじめ定められたレベルに正規化され、帯域制限された雑音信号をフレーム長に等しい時間長だけ発生させ、ゲイン回路１４０に出力する。この実施の形態では、雑音信号として、一例として白色雑音を用いるが、他の雑音信号を使用してもよい。
【００２８】
有声/無声判別回路２００は、狭帯域入力信号x(n)を入力し、フレーム毎の信号が有声か無声かを判別する。有声/無声の判定として、例えば、狭帯域入力信号x(n)に対して、あらかじめ定められた遅れ時間ｍまでの正規化自己相関関数D（T）を式（１）に従って計算し、D(T)の最大値を求め、D(T)の最大値があらかじめ定められたしきい値より大きければ有声、さもなければ無声と判別する。
【００２９】

【００３０】
そして、有声/無声判別回路２００は、有声/無声判別情報をゲイン調整回路２１０に出力する。なお式(1)において、Nは正規化自己相関を計算するためのサンプル数である。
【００３１】
ゲイン調整回路３１０は、有声/無声判別回路２００から有声/無声判別情報を入力し、有声/無声に応じて、雑音信号に与えるゲインを調整し、ゲイン回路１４０に出力する。
【００３２】
ゲイン回路１４０は、ゲイン調整回路３１０からゲインを入力し、雑音発生回路１２０の出力信号にゲインを乗じて、合成フィルタ回路１７０に出力する。
【００３３】
合成フィルタ回路１７０は、ゲイン回路１４０からの出力信号を入力し、さらに係数計算部１３０から、あらかじめ定められた次数の係数を入力してフィルタを構成し、帯域拡張化に必要な高周波域信号y(n)を出力する。
【００３４】
標本化周波数変換回路１８０は、狭帯域入力信号x(n)を、あらかじめ定められた標本化周波数にアップサンプリングし、アップサンプリング後の信号s(n)を出力する。
【００３５】
加算器１９０は、合成フィルタ回路１７０の出力信号y(n)と、標本化周波数変換回路１８０の出力信号s(n)を加算し、最終的に帯域拡張された信号z(n)を形成して出力する。
【００３６】
第２の符号化回路１９５は、加算器１９０の出力信号z(n)を入力し、第２の符号化方式により符号化を行い第２の符号を得て出力する。
【００３７】
以上で第１の実施の形態の説明を終える。
【００３８】
図２は、本発明の第２の実施の形態の構成を示すブロックである。図２を参照すると、本発明の第２の実施の形態のトランスコーダは、第１の復号回路１０５と、スペクトルパラメータ計算回路１００と、適応コードブック回路１１０と、雑音発生回路１２０と、係数計算回路１３０と、ゲイン回路３４０と、合成フィルタ回路１７０と、標本化周波数変換回路１８０と、加算器１６０と、加算器１９０と、第２の符号化回路１９５と、有声/無声判別回路２００と、ゲイン調整回路２１０と、を備えている。図２において、図１と同一の要素には、同一の参照符号が付されている。以下では、主に、前記第１の実施の形態との相違点について説明し、図１と同一の要素の説明は、適宜省略する。図２を参照すると、本発明の第２の実施の形態は、図１の構成に加え、適応コードブック回路１１０と、加算器１６０を備えている。
【００３９】
有声/無声判別回路２００は、狭帯域入力信号x(n)を入力し、フレーム毎の信号が有声か無声かを判別する。有声/無声の判定として、例えば、狭帯域入力信号x(n)に対して、あらかじめ定められた遅れ時間ｍまでの正規化自己相関関数D（T）を上式（１）に従って計算し、D(T)の最大値を求め、D(T)の最大値があらかじめ定められたしきい値より大きければ有声、さもなければ無声と判別し、判別結果をゲイン調整回路２１０へ出力する。
【００４０】
また、有声/無声判別回路２００は、有声部分のフレームでは、正規化自己相関関数D（T）を最大化するＴの値をピッチ周期Ｔとして適応コードブック回路１１０へ供給する。
【００４１】
適応コードブック回路１１０は、適応コードブックの遅延Ｔを、有声/無声判別回路２００から入力し、過去の音源信号v(n)をもとに、次式（２）に従い適応コードベクトルp(n)を発生し出力する。
【００４２】

【００４３】
ゲイン調整回路２１０は、有声/無声判別回路２００から有声/無声判別情報を入力し、有声か無声かに応じて、適応コードブック信号のゲインと、雑音信号のゲインを調整し、ゲイン回路３４０に供給する。
【００４４】
ゲイン回路３４０は、ゲイン調整回路２１０からゲインを入力し、適応コードブック回路１１０と雑音発生回路１２０の少なくとも一方の出力信号に、ゲインを乗じて加算器１６０に出力する。
【００４５】
加算器１６０は、ゲイン回路３４０から出力される２種類の信号（適応コードブック回路１１０と雑音発生回路１２０の少なくとも一方の出力信号にゲインを乗じた２つの信号）を加算して、合成フィルタ回路１７０と適応コードブック回路１１０とに出力する。
【００４６】
合成フィルタ回路１７０は、加算器１６０の出力信号を入力し、さらに係数計算部１３０から、あらかじめ定められた次数の係数（フィルタ係数）を入力してフィルタを構成し、帯域拡張に必要な高周波域の信号y(n)を出力する。
【００４７】
本発明の第２の実施の形態によれば、高い周波数部分の過去の音源信号をもとに、狭帯域入力信号から計算した遅延を用いて適応コードブック信号を発生させ適切なゲインを乗じて雑音信号と加算することにより、母音などのように、高い周波数部分の信号に周期性が必要な場合に音質の良好な帯域拡張信号を発生できる。以上で第２の実施の形態の説明を終える。
【００４８】
なお、本発明の第２の実施の形態の変形として、図２の適応コードブック回路１１０のかわりに、図６に示すように、周期信号発生回路１１５を備えた構成としてもよい。周期信号発生回路１１５は、有声/無声判別回路２００からピッチ周期を入力し、該ピッチ周期に用いて、周期信号を発生させてゲイン回路３４０に出力する。周期信号発生回路１１５以外の構成は、前記第２の実施の形態と同様である。
【００４９】
図３は、本発明の第３の実施の形態の構成を示すブロック図である。図３を参照すると、本発明の第３の実施の形態のトランスコーダは、第１の復号回路１０５と、スペクトルパラメータ計算回路１００と、適応コードブック回路１１０と、雑音発生回路１２０と、係数計算回路１３０と、ゲイン回路３００と、合成フィルタ回路１７０と、標本化周波数変換回路１８０と、加算器１９０と、第２の符号化回路１９５と、有声/無声判別回路２００と、ゲイン調整回路２１０と、ピッチプリフィルタ回路４００と、を備えている。図３において、図１、図２と同一又は同等の要素には、同一の参照符号が付されている。以下では、主に、第２の実施の形態との相違点について説明し、図１、図２と同一の要素の説明は省略する。本実施の形態では、ピッチプリフィルタ回路４００を備えている。
【００５０】
ゲイン回路３００は、ゲイン調整回路２１０からゲインを入力し、適応コードブック回路１１０と雑音発生回路１２０の出力信号にゲインを乗じて２種類の信号を加算し加算結果を、ピッチプリフィルタ回路４００に出力する。
【００５１】
ピッチプリフィルタ回路４００は、有声/無声判別回路２００から遅延Ｔ（ピッチ周期）を入力し、ゲイン回路３００からの音源信号v(n)に対し、次式（３）に従い、ピッチプリフィルタリングを行った上で、合成フィルタ回路１７０に出力する。
【００５２】

【００５３】
この実施の形態によれば、遅延を利用して音源信号に対しピッチプリフィルタ回路４００を用いることで、良好な音質の帯域拡張信号を発生することができる。以上で第３の実施の形態の説明を終える。
【００５４】
なお、本実施の形態においても、前記第２の実施の形態の変形例と同様、適応コードブック回路１１０のかわりに、周期信号発生回路を用いて構成してもよい。この場合は、周期信号発生回路は、有声/無声判別回路２００から信号を入力してピッチ周期を計算し、ピッチ周期に基づく周期信号を発生させ、ゲイン回路３００に出力する。
【００５５】
図４は、本発明の第４の実施の形態の構成を示すブロック図である。図４を参照すると、本発明の第４の実施の形態のトランスコーダは、第１の復号回路１０５と、スペクトルパラメータ計算回路１００と、適応コードブック回路１１０と、雑音発生回路１２０と、係数計算回路１３０と、ゲイン回路３４０と、加算器１６０と、合成フィルタ回路１７０と、標本化周波数変換回路１８０と、加算器１９０と、第２の符号化回路１９５と、有声/無声判別回路２００と、ゲイン調整回路２１０と、低域通過フィルタ回路５００とを備えている。図４において、図２と同一又は同等の要素には、同一の参照符号が付されている。以下では、主に前記第２の実施の形態との相違点について説明し、図２と同一の要素の説明は省略する。図４を参照すると、適応コードブック回路１１０の出力を入力とする低域通過フィルタ回路５００を備えている。
【００５６】
低域通過フィルタ（LPF）回路５００は、適応コードブック回路１１０の出力信号に対して、式（４）により、低域信号を通過させた上で、ゲイン回路３４０に出力する。
【００５７】
p’(n)=p(n)*h(n) (4)
【００５８】
低域通過フィルタ回路５００のカットオフ周波数は、あらかじめ定めておき、例えば、６kHzとすることができる。なお、式(4)で、h(n)は低域通過フィルタのインパルス応答を、記号”*”は畳み込み演算を、それぞれ示す。
【００５９】
以上で本発明の第４の実施の形態の説明を終える。なお、本発明の第４の実施の形態においても、前記第２の実施の形態の変形例と同様、適応コードブック回路１１０のかわりに、周期信号発生回路を用いることもできる。この場合は、有声/無声判別回路２００から信号を入力してピッチ周期を計算し、ピッチ周期に基づく周期信号を発生させ、ゲイン回路３４０に出力する。
【００６０】
図５は、本発明の第５の実施の形態の構成を示すブロック図である。図５を参照すると、本発明の第５の実施の形態のトランスコーダは、第１の復号回路１０５と、スペクトルパラメータ計算回路１００と、適応コードブック回路１１０と、雑音発生回路１２０と、係数計算回路１３０と、ゲイン回路３００と、合成フィルタ回路１７０と、標本化周波数変換回路１８０と、加算器１９０と、第２の符号化回路１９５と、有声/無声判別回路２００と、ゲイン調整回路２１０と、ピッチプリフィルタ４００と、ポストフィルタ６００と、を備えている。図５において、図３と同一又は同等の要素には、同一の参照符号が付されている。以下では、主に、前記第３の実施の形態との相違点について説明し、図３と同一の要素の説明は省略する。本実施の形態では、前記第３の実施の形態の構成に加えて、ポストフィルタ６００をさらに備えている。
【００６１】
ポストフィルタ６００は、係数計算回路１３０から係数（フィルタ係数）を入力し、係数に重み付けを施した上で、式(5)に従い、ポストフィルタリングを行い、出力を加算器１９０に出力する。
【００６２】
y'(n)=y(n) - Σａ_ｉγ₁ ^ｉy(n-i) + Σａ_ｉγ₂ ^ｉy'(n-i) (5)
【００６３】
ポストフィルタ６００を用いることにより、良好な音質の帯域拡張信号を発生することができる。以上で第５の実施の形態の説明を終える。
【００６４】
なお、本発明の第５の実施の形態においても、前記第２の実施の形態の変形例と同様、適応コードブック回路１１０のかわりに、周期信号発生回路を用いることもできる。この場合は、有声/無声判別回路２００から信号を入力してピッチ周期を計算し、ピッチ周期に基づく周期信号を発生させ、ゲイン回路３００に出力する。
【００６５】
上記第５の実施の形態で説明したポストフィルタを前記第１の実施の形態に用いる等、各実施の形態の構成を組み合わせてもよい。以上本発明を上記各実施の形態に即して説明したが、本発明は、上記実施の形態の構成にのみ限定されるものでなく、特許請求の範囲の各請求項の発明の範囲内で、当業者であればなし得るであろう各種変形、修正を含むことは勿論である。
【００６６】
【発明の効果】
以上説明したように、本発明のトランスコーダによれば、第１の符号化方式による符号を入力し第２の符号化方式による符号に変換して出力する際に、信号の帯域を拡張化した上で変換し、比較的演算量の少ない処理により高い周波数の信号を発生させ、狭帯域入力信号の標本化周波数を変換させた信号と加算させることにより、帯域拡張化信号（例えば７kHz帯域）を発生させる、ことができるという効果がある。
【００６７】
また、本発明のトランスコーダによれば、高い周波数部分の過去の音源信号をもとに、狭帯域入力信号から計算した遅延を用いて適応コードブック信号を発生させ、適切なゲインを乗じて雑音信号と加算することにより、母音などのように、高い周波数部分の信号に周期性が必要な場合に音質の良好な帯域拡張信号を発生させることができる、という効果がある。
【００６８】
さらに、本発明によれば、遅延を利用して音源信号に対しピッチプリフィルタを用いるか、あるいは、係数計算回路からの係数に重み付けをしてポストフィルタを備える構成とすることで、さらに良好な音質の帯域拡張信号を発生することができる、という効果がある。
【図面の簡単な説明】
【図１】本発明の第１の実施の形態の構成を示す図である。
【図２】本発明の第２の実施の形態の構成を示す図である。
【図３】本発明の第３の実施の形態の構成を示す図である。
【図４】本発明の第４の実施の形態の構成を示す図である。
【図５】本発明の第５の実施の形態の構成を示す図である。
【図６】本発明の第２の実施の形態の変形例の構成を示す図である。
【符号の説明】
１００スペクトルパラメータ計算回路
１０５第１の復号回路（第１の復号化部）
１１０適応コードブック回路
１１５周期信号発生回路
１２０雑音発生回路
１３０係数計算回路
１４０ゲイン回路
１６０、１９０加算器
１７０合成フィルタ回路
１８０標本化周波数変換回路
１９５第２の符号化回路（第２の符号化部）
２００有声／無声判別回路
２１０、３１０ゲイン調整回路
３００、３４０ゲイン回路
４００ピッチプリフィルタ回路
５００低域通過フィルタ回路
６００ポストフィルタ回路[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a transcoder that mutually converts a code according to a first encoding method and a code according to a second encoding method, and converts a frequency band of a signal when converting from a first code to a second code. The present invention relates to a transcoder that performs conversion after expansion.
[0002]
[Prior art]
A method is known in which the frequency band of an audio signal encoded and reproduced at a low bit rate is expanded on the receiving side without transmitting auxiliary information for band expansion from the transmitting side (for example, Non-Patent Document 1). .
[0003]
[Non-Patent Document 1]
P.Jax, P.Vary, "Wideband extension of telephone speech using hidden markov model," Proc. IEEE Speech Coding Workshop, pp.133-135, 2000 [0004]
In this conventional method, a filter coefficient after band expansion using an HMM (Hidden Markov Model) is searched for on the receiving side.
[0005]
On the other hand, in the transcoder that mutually converts the code according to the first encoding method and the code according to the second encoding method, the band of the signal is expanded when converting from the first code to the second code. There is no precedent for the transcoder to convert above.
[0006]
[Problems to be solved by the invention]
In the conventional method described in the above-mentioned document by P.Jax and P.Vary (Non-Patent Document 1 above), it is necessary to model the spectral envelope of wideband speech and the filter coefficient by HMM. there were.
[0007]
That is, it is necessary to determine parameters of the HMM model in advance offline from a large amount of speech databases, which requires a lot of calculation time and cost.
[0008]
In addition, when performing bandwidth expansion processing in real time on the receiving side, a search using an HMM model is necessary, which requires a large amount of computation.
[0009]
Therefore, the main object of the present invention is to convert a code according to the first encoding scheme and a code according to the second encoding scheme from each other into a second code, when converting from the first code to the second code. It is an object of the present invention to provide a transcoder and a code conversion method capable of performing band expansion conversion with a relatively small amount of calculation and good sound quality when performing conversion after extending a signal band.
[0010]
[Means for Solving the Problems]
The present invention that achieves the above object is a transcoder for interconverting a code according to a first encoding scheme and a code according to a second encoding scheme, and according to the first aspect, the first encoding scheme A spectrum parameter calculating unit that inputs a code encoded by a method and calculates a spectrum parameter that is decoded by the first encoding method and represents a spectrum characteristic; a noise generating unit that generates a noise signal; and a frequency of the spectrum parameter A coefficient calculation unit for obtaining a filter coefficient after shifting the signal, a gain unit for giving an appropriate gain to the output of the noise generation unit, and an output of the gain unit through a synthesis filter configured using the coefficient to expand the band A synthesis filter unit that reproduces a signal, and an output signal obtained by adding the output signal of the synthesis filter unit after converting the sampling frequency of the input signal. And parts, are configured to output a second code by coding a by the output signal of the adding unit second encoding method.
[0011]
According to the second aspect, the transcoder according to the present invention is a transcoder that mutually converts a code according to the first encoding scheme and a code according to the second encoding scheme. A spectral parameter calculation unit that inputs a code encoded by the first encoding method, calculates a spectral parameter that represents a spectral characteristic by decoding using the first encoding method, calculates a pitch period from the input signal, and calculates the pitch period and past sound sources An adaptive codebook unit that generates an adaptive codebook component based on a signal, a noise generation unit that generates a noise signal, a coefficient calculation unit that obtains a filter coefficient after shifting the frequency of the spectral parameter, and the noise generation unit Are added after an appropriate gain is given to at least one of the output of the adaptive codebook and the output of the adaptive codebook section, and a sound source signal is output. A synthesis unit configured to input the sound source signal to a synthesis filter configured using the coefficients and reproduce the band extension signal, and convert the sampling frequency of the reproduction signal and then output the synthesis filter unit An adder for adding and outputting signals, and an output signal of the adder are encoded by a second encoding method and a second code is output.
[0012]
According to the third aspect, the transcoder according to the present invention is a transcoder that mutually converts a code according to the first encoding scheme and a code according to the second encoding scheme. A spectral parameter calculation unit that inputs a code encoded by the first encoding method, calculates a spectral parameter that represents a spectral characteristic by decoding using the first encoding method, calculates a pitch period from the input signal, and calculates the pitch period and past sound sources An adaptive codebook unit that generates an adaptive codebook component based on a signal, a noise generation unit that generates a noise signal, a coefficient calculation unit that obtains a filter coefficient after shifting the frequency of the spectral parameter, and the noise generation unit Are added after an appropriate gain is given to at least one of the output of the adaptive codebook and the output of the adaptive codebook section, and a sound source signal is output. And a synthesizing filter unit that inputs a pitch prefilter output signal to a synthesizing filter configured using the coefficients and reproduces a band extension signal by passing a pitch prefilter through the sound source signal using the pitch period. An addition unit that converts the sampling frequency of the reproduction signal and then adds and outputs the output signal of the synthesis filter unit; and an output signal of the addition unit is encoded by a second encoding method to be second. The code is output.
[0013]
Further, according to the present invention, the output of the adaptive codebook unit may be passed through a low-pass filter having a predetermined cutoff frequency.
[0014]
Further, according to the present invention, a post filter may be configured using a weighting coefficient obtained by weighting the coefficient, and a band extension signal may be reproduced by passing the output signal of the synthesis filter unit through the post filter. .
[0015]
According to one aspect of the method of the present invention, in the transcoder code conversion method for mutually converting the code according to the first encoding method and the code according to the second encoding method, the first encoding method is used. Decoding the code encoded by the first decoding scheme and outputting a decoded signal;
Calculating and outputting spectral parameters representing spectral characteristics from the decoded signal;
Obtaining and outputting a filter coefficient after shifting the frequency of the spectral parameter;
Giving a gain to the output signal from the noise generator;
Passing the output signal given the gain to the synthesis filter configured using the filter coefficient, and outputting a signal in a band necessary for band conversion;
Adding a signal obtained by converting the decoded signal at a predetermined sampling frequency and an output signal of the synthesis filter;
Encoding the addition result by the second encoding method and obtaining and outputting a second code.
[0016]
Alternatively, according to another aspect of the method according to the present invention, the step of decoding the code encoded by the first encoding scheme by the first decoding scheme and outputting a decoded signal;
Calculating and outputting spectral parameters representing spectral characteristics from the decoded signal;
Calculating a pitch period from the decoded signal and generating an adaptive codebook component based on the pitch period and a past sound source signal;
Obtaining and outputting a filter coefficient after shifting the frequency of the spectral parameter;
Adding a gain to at least one of the output of the noise generation unit and the adaptive codebook component and outputting the sound source signal; and
Outputting a signal of a band necessary for band conversion through the sound source signal to a synthesis filter configured using the filter coefficient;
Adding a signal obtained by converting the decoded signal at a predetermined sampling frequency and an output signal of the synthesis filter;
Encoding the addition result by the second encoding method and obtaining and outputting a second code.
[0017]
Alternatively, according to another aspect of the method according to the present invention, the step of decoding the code encoded by the first encoding scheme by the first decoding scheme and outputting a decoded signal;
Calculating and outputting spectral parameters representing spectral characteristics from the decoded signal;
Calculating a pitch period from the decoded signal and generating an adaptive codebook component based on the pitch period and a past sound source signal;
Obtaining and outputting a filter coefficient after shifting the frequency of the spectral parameter;
Adding a gain to at least one of the output of the noise generation unit and the adaptive codebook component and outputting the sound source signal; and
Performing a pitch pre-filter on the sound source signal using the pitch period;
Outputting a signal of a band necessary for band conversion through the pitch prefiltered signal to a synthesis filter configured using the filter coefficient;
Adding a signal obtained by converting the decoded signal at a predetermined sampling frequency and an output signal of the synthesis filter;
Encoding the addition result by the second encoding method and obtaining and outputting a second code.
[0018]
According to another aspect of the method of the present invention, the pitch pre-filter uses the pitch period to perform pre-filter processing on the sound source signal from the gain unit, and outputs to the synthesis filter from the pitch pre-filter. It is good also as a structure which has the step which inputs a signal.
[0019]
According to another aspect of the method of the present invention, the step of inputting the output signal of the synthesis filter unit to a post filter configured using a weighting coefficient obtained by weighting the filter coefficient from the counting calculation unit. It is good also as a structure to have.
[0020]
Furthermore, according to another aspect of the method of the present invention, instead of the output signal from the adaptive codebook unit, the output of the periodic signal generation unit that generates the periodic signal using the pitch period is supplied to the gain unit. May be.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described with reference to the drawings. In the following, the first code is encoded with respect to a 4 kHz band narrowband input signal. In the transcoder, the second code is expanded to a signal of 5 kHz band or 7 kHz band, It is assumed that the second code is obtained by encoding with the encoding method.
[0022]
FIG. 1 is a block diagram showing a configuration of a first embodiment of a transcoder according to the present invention. Referring to FIG. 1, the transcoder includes a first decoding circuit 105, a spectrum parameter calculation circuit 100, a noise generation circuit 120, a coefficient calculation circuit 130, a synthesis filter circuit 170, and a sampling frequency conversion circuit 180. , An adder 190, a second encoding circuit 195, a voiced / unvoiced discrimination circuit 200, a gain adjustment circuit 310, and a gain circuit 140.
[0023]
The first decoding circuit 105 receives a code according to the first encoding method, decodes it according to the first decoding method, and outputs a decoded signal x (n).
[0024]
The spectrum parameter calculation circuit 100 divides the decoded signal x (n) into frames (for example, 10 ms), and calculates a spectrum parameter of a predetermined order P for each frame. Here, the spectrum parameter is a parameter representing the outline of the spectrum of the audio signal for each frame, and a known LPC (Linear Predictive Coding) analysis or the like can be used for this calculation. Further, the spectral parameter calculation unit 100 converts the linear prediction coefficient αi (i = 1,... P) calculated by the LPC analysis into an LSP (Line Spectrum Pair) parameter suitable for quantization and interpolation. Output. Here, for conversion from the linear prediction coefficient to the LSP, for example, the following paper is referred to (Non-Patent Document 2).
[0025]
[Non-Patent Document 2]
Kashimura, Itakura: “Speech information compression by line spectrum pair (LSP) speech analysis and synthesis method”, IEICE Transactions, J64-A, pp.599-606, 1981 [0026]
The coefficient calculation circuit 130 receives the spectrum parameter output from the spectrum parameter calculation circuit 100 and converts the spectrum parameter into a band-extended signal coefficient. For this conversion, for example, a known method such as a method of simply shifting the LSP frequency to a higher frequency, a non-linear conversion method, or a linear conversion method can be used. Here, using all or part of the LSP parameters, the existing frequency band of the LSP is shifted to a higher frequency band, and then converted into a linear prediction coefficient of order P, and output to the synthesis filter circuit 170.
[0027]
The noise generator 120 generates a noise signal whose average amplitude is normalized to a predetermined level and is band-limited for a time length equal to the frame length, and outputs the generated noise signal to the gain circuit 140. In this embodiment, white noise is used as an example of the noise signal, but other noise signals may be used.
[0028]
Voiced / unvoiced discrimination circuit 200 receives narrowband input signal x (n) and discriminates whether the signal for each frame is voiced or unvoiced. For the voiced / unvoiced determination, for example, a normalized autocorrelation function D (T) up to a predetermined delay time m is calculated for the narrowband input signal x (n) according to the equation (1), and D ( The maximum value of T) is obtained, and if the maximum value of D (T) is larger than a predetermined threshold value, it is determined that the voice is voiced, and otherwise, it is discriminated.
[0029]

[0030]
Voiced / unvoiced discrimination circuit 200 outputs voiced / unvoiced discrimination information to gain adjustment circuit 210. In Equation (1), N is the number of samples for calculating the normalized autocorrelation.
[0031]
The gain adjustment circuit 310 receives the voiced / unvoiced discrimination information from the voiced / unvoiced discrimination circuit 200, adjusts the gain applied to the noise signal according to voiced / unvoiced, and outputs it to the gain circuit 140.
[0032]
The gain circuit 140 receives the gain from the gain adjustment circuit 310, multiplies the output signal of the noise generation circuit 120 by the gain, and outputs the result to the synthesis filter circuit 170.
[0033]
The synthesis filter circuit 170 receives the output signal from the gain circuit 140, and further inputs a coefficient of a predetermined order from the coefficient calculation unit 130 to form a filter, and the high frequency band signal y required for band expansion. Output (n).
[0034]
The sampling frequency conversion circuit 180 upsamples the narrowband input signal x (n) to a predetermined sampling frequency, and outputs a signal s (n) after the upsampling.
[0035]
The adder 190 adds the output signal y (n) of the synthesis filter circuit 170 and the output signal s (n) of the sampling frequency conversion circuit 180, and finally forms a band-extended signal z (n). Output.
[0036]
The second encoding circuit 195 receives the output signal z (n) of the adder 190, performs encoding by the second encoding method, and obtains and outputs the second code.
[0037]
This is the end of the description of the first embodiment.
[0038]
FIG. 2 is a block diagram showing the configuration of the second exemplary embodiment of the present invention. Referring to FIG. 2, the transcoder according to the second embodiment of the present invention includes a first decoding circuit 105, a spectral parameter calculation circuit 100, an adaptive codebook circuit 110, a noise generation circuit 120, and a coefficient calculation. A circuit 130, a gain circuit 340, a synthesis filter circuit 170, a sampling frequency conversion circuit 180, an adder 160, an adder 190, a second encoding circuit 195, a voiced / unvoiced discrimination circuit 200, A gain adjustment circuit 210. In FIG. 2, the same elements as those in FIG. 1 are denoted by the same reference numerals. Hereinafter, differences from the first embodiment will be mainly described, and description of the same elements as those in FIG. 1 will be omitted as appropriate. Referring to FIG. 2, the second embodiment of the present invention includes an adaptive codebook circuit 110 and an adder 160 in addition to the configuration of FIG.
[0039]
Voiced / unvoiced discrimination circuit 200 receives narrowband input signal x (n) and discriminates whether the signal for each frame is voiced or unvoiced. For the determination of voiced / unvoiced, for example, a normalized autocorrelation function D (T) up to a predetermined delay time m is calculated according to the above equation (1) for a narrowband input signal x (n), and D The maximum value of (T) is obtained, and if the maximum value of D (T) is larger than a predetermined threshold value, it is determined as voiced, otherwise it is determined as unvoiced, and the determination result is output to the gain adjustment circuit 210.
[0040]
In addition, the voiced / unvoiced discrimination circuit 200 supplies a value of T that maximizes the normalized autocorrelation function D (T) to the adaptive codebook circuit 110 as the pitch period T in the voiced frame.
[0041]
The adaptive codebook circuit 110 inputs the adaptive codebook delay T from the voiced / unvoiced discrimination circuit 200, and based on the past sound source signal v (n), the adaptive code vector p (n ) Is generated and output.
[0042]

[0043]
The gain adjustment circuit 210 receives the voiced / unvoiced discrimination information from the voiced / unvoiced discrimination circuit 200, adjusts the gain of the adaptive codebook signal and the noise signal according to whether it is voiced or unvoiced, and sends it to the gain circuit 340. Supply.
[0044]
The gain circuit 340 receives the gain from the gain adjustment circuit 210, multiplies the output signal of at least one of the adaptive codebook circuit 110 and the noise generation circuit 120 by the gain, and outputs the result to the adder 160.
[0045]
The adder 160 adds two types of signals output from the gain circuit 340 (two signals obtained by multiplying at least one of the output signals of the adaptive codebook circuit 110 and the noise generation circuit 120 by a gain) to generate a synthesis filter circuit. 170 and the adaptive code book circuit 110.
[0046]
The synthesis filter circuit 170 receives the output signal of the adder 160, and further inputs a predetermined order coefficient (filter coefficient) from the coefficient calculation unit 130 to form a filter, and a high frequency band necessary for band expansion. The signal y (n) is output.
[0047]
According to the second embodiment of the present invention, an adaptive codebook signal is generated using a delay calculated from a narrowband input signal based on a past sound source signal of a high frequency portion and multiplied by an appropriate gain. By adding to the noise signal, it is possible to generate a band extension signal with good sound quality when periodicity is required for a signal in a high frequency part such as a vowel. This is the end of the description of the second embodiment.
[0048]
As a modification of the second embodiment of the present invention, a configuration including a periodic signal generation circuit 115 as shown in FIG. 6 instead of the adaptive codebook circuit 110 of FIG. 2 may be adopted. The periodic signal generation circuit 115 receives the pitch period from the voiced / unvoiced discrimination circuit 200, generates a periodic signal using the pitch period, and outputs the periodic signal to the gain circuit 340. The configuration other than the periodic signal generation circuit 115 is the same as that of the second embodiment.
[0049]
FIG. 3 is a block diagram showing the configuration of the third exemplary embodiment of the present invention. Referring to FIG. 3, the transcoder according to the third embodiment of the present invention includes a first decoding circuit 105, a spectral parameter calculation circuit 100, an adaptive codebook circuit 110, a noise generation circuit 120, a coefficient calculation. Circuit 130, gain circuit 300, synthesis filter circuit 170, sampling frequency conversion circuit 180, adder 190, second encoding circuit 195, voiced / unvoiced discrimination circuit 200, gain adjustment circuit 210, And a pitch pre-filter circuit 400. In FIG. 3, the same or equivalent elements as those in FIGS. 1 and 2 are denoted by the same reference numerals. Hereinafter, differences from the second embodiment will be mainly described, and description of the same elements as those in FIGS. 1 and 2 will be omitted. In the present embodiment, a pitch prefilter circuit 400 is provided.
[0050]
The gain circuit 300 receives the gain from the gain adjustment circuit 210, multiplies the output signals of the adaptive codebook circuit 110 and the noise generation circuit 120 by the gain, adds two types of signals, and adds the addition result to the pitch prefilter circuit 400. Output.
[0051]
The pitch prefilter circuit 400 receives the delay T (pitch period) from the voiced / unvoiced discrimination circuit 200 and performs pitch prefiltering on the sound source signal v (n) from the gain circuit 300 according to the following equation (3). Then, it is output to the synthesis filter circuit 170.
[0052]

[0053]
According to this embodiment, by using the pitch pre-filter circuit 400 for the sound source signal using the delay, a band extension signal with good sound quality can be generated. This is the end of the description of the third embodiment.
[0054]
In the present embodiment, a periodic signal generation circuit may be used instead of the adaptive code book circuit 110, as in the modification of the second embodiment. In this case, the periodic signal generation circuit inputs a signal from the voiced / unvoiced discrimination circuit 200, calculates the pitch period, generates a periodic signal based on the pitch period, and outputs it to the gain circuit 300.
[0055]
FIG. 4 is a block diagram showing the configuration of the fourth exemplary embodiment of the present invention. Referring to FIG. 4, the transcoder according to the fourth embodiment of the present invention includes a first decoding circuit 105, a spectral parameter calculation circuit 100, an adaptive codebook circuit 110, a noise generation circuit 120, and a coefficient calculation. A circuit 130, a gain circuit 340, an adder 160, a synthesis filter circuit 170, a sampling frequency conversion circuit 180, an adder 190, a second encoding circuit 195, a voiced / unvoiced discrimination circuit 200, A gain adjustment circuit 210 and a low-pass filter circuit 500 are provided. In FIG. 4, the same or equivalent elements as those in FIG. Hereinafter, differences from the second embodiment will be mainly described, and description of the same elements as those in FIG. 2 will be omitted. Referring to FIG. 4, a low-pass filter circuit 500 that receives the output of the adaptive codebook circuit 110 is provided.
[0056]
The low-pass filter (LPF) circuit 500 passes the low-frequency signal to the gain circuit 340 after passing the low-frequency signal with respect to the output signal of the adaptive codebook circuit 110 according to Expression (4).
[0057]
p '(n) = p (n) * h (n) (4)
[0058]
The cut-off frequency of the low-pass filter circuit 500 is predetermined and can be set to 6 kHz, for example. In equation (4), h (n) represents the impulse response of the low-pass filter, and symbol “*” represents the convolution operation.
[0059]
This is the end of the description of the fourth embodiment of the present invention. In the fourth embodiment of the present invention, a periodic signal generation circuit can be used instead of the adaptive codebook circuit 110, as in the modification of the second embodiment. In this case, a signal is input from voiced / unvoiced discrimination circuit 200 to calculate the pitch period, a periodic signal based on the pitch period is generated, and output to gain circuit 340.
[0060]
FIG. 5 is a block diagram showing the configuration of the fifth exemplary embodiment of the present invention. Referring to FIG. 5, the transcoder according to the fifth embodiment of the present invention includes a first decoding circuit 105, a spectral parameter calculation circuit 100, an adaptive codebook circuit 110, a noise generation circuit 120, and coefficient calculation. Circuit 130, gain circuit 300, synthesis filter circuit 170, sampling frequency conversion circuit 180, adder 190, second encoding circuit 195, voiced / unvoiced discrimination circuit 200, gain adjustment circuit 210, The pitch pre-filter 400 and the post filter 600 are provided. In FIG. 5, the same or equivalent elements as those in FIG. 3 are denoted by the same reference numerals. In the following, differences from the third embodiment will be mainly described, and description of the same elements as those in FIG. 3 will be omitted. In the present embodiment, a post filter 600 is further provided in addition to the configuration of the third embodiment.
[0061]
The post filter 600 receives a coefficient (filter coefficient) from the coefficient calculation circuit 130, weights the coefficient, performs post filtering according to Equation (5), and outputs the output to the adder 190.
[0062]
y '(n) = y ( n) - Σa i γ 1 i y (ni) + Σa i γ 2 i y' (ni) (5)
[0063]
By using the post filter 600, it is possible to generate a band extension signal with good sound quality. This is the end of the description of the fifth embodiment.
[0064]
In the fifth embodiment of the present invention, a periodic signal generation circuit can be used in place of the adaptive codebook circuit 110, as in the modification of the second embodiment. In this case, a signal is input from the voiced / unvoiced discrimination circuit 200 to calculate the pitch period, and a periodic signal based on the pitch period is generated and output to the gain circuit 300.
[0065]
You may combine the structure of each embodiment, such as using the post filter demonstrated in the said 5th Embodiment for the said 1st Embodiment. Although the present invention has been described with reference to each of the above embodiments, the present invention is not limited only to the configuration of the above embodiments, and is within the scope of the invention of each claim. It goes without saying that various modifications and corrections that can be made by those skilled in the art are included.
[0066]
【The invention's effect】
As described above, according to the transcoder of the present invention, when the code according to the first encoding method is input and converted into the code according to the second encoding method and output, the signal band is expanded. By converting the above, generating a signal with a high frequency by processing with a relatively small amount of computation, and adding the sampling frequency of the narrowband input signal to the converted signal, a band expansion signal (for example, 7 kHz band) is obtained. There is an effect that it can be generated.
[0067]
Further, according to the transcoder of the present invention, an adaptive codebook signal is generated using a delay calculated from a narrowband input signal based on a past sound source signal of a high frequency portion, and multiplied by an appropriate gain to generate noise. By adding it to the signal, there is an effect that it is possible to generate a band extension signal with good sound quality when periodicity is required for a signal of a high frequency portion such as a vowel.
[0068]
Furthermore, according to the present invention, a pitch prefilter is used for the sound source signal using a delay, or the coefficient from the coefficient calculation circuit is weighted, and the post filter is provided. There is an effect that a sound quality band extension signal can be generated.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration of a first exemplary embodiment of the present invention.
FIG. 2 is a diagram showing a configuration of a second exemplary embodiment of the present invention.
FIG. 3 is a diagram showing a configuration of a third exemplary embodiment of the present invention.
FIG. 4 is a diagram showing a configuration of a fourth exemplary embodiment of the present invention.
FIG. 5 is a diagram showing a configuration of a fifth exemplary embodiment of the present invention.
FIG. 6 is a diagram showing a configuration of a modification of the second embodiment of the present invention.
[Explanation of symbols]
100 spectrum parameter calculation circuit 105 first decoding circuit (first decoding unit)
110 Adaptive Codebook Circuit 115 Periodic Signal Generation Circuit 120 Noise Generation Circuit 130 Coefficient Calculation Circuit 140

Gain Circuit

160, 190 Adder 170 Synthesis Filter Circuit 180 Sampling Frequency Conversion Circuit 195 Second Encoding Circuit (Second Encoding Unit) )
200 Voiced /

Unvoiced Discriminating Circuit

210, 310

Gain Adjustment Circuit

300, 340 Gain Circuit 400 Pitch Pre-Filter Circuit 500 Low-Pass Filter Circuit 600 Post-Filter Circuit

Claims

In a transcoder that mutually converts a code according to the first encoding scheme and a code according to the second encoding scheme,
A first decoding unit that inputs a code encoded by the first encoding method, decodes by the first decoding method, and outputs a decoded signal;
A spectral parameter calculation unit that inputs the decoded signal and calculates and outputs a spectral parameter representing spectral characteristics;
A noise generator for generating a noise signal;
A coefficient calculation unit that shifts the frequency of the spectral parameter to a high frequency and then converts it to a filter coefficient and outputs it;
A gain unit that gives a gain to the output signal from the noise generation unit and outputs the gain,
A synthesis filter having a synthesis filter configured by using the filter coefficient from the coefficient calculation unit, inputting an output signal from the gain unit, passing the synthesis signal through the synthesis filter, and outputting a signal in a band necessary for band conversion And
A sampling frequency conversion circuit that outputs a signal obtained by converting the decoded signal at a predetermined sampling frequency;
An adder that adds and outputs the output signal of the sampling frequency conversion circuit and the output signal of the synthesis filter;
A second encoding unit that inputs an output signal of the adder, performs encoding by the second encoding method, and obtains and outputs a second code;
A transcoder characterized by comprising:

In a transcoder that mutually converts a code according to the first encoding scheme and a code according to the second encoding scheme,
A first decoding unit that inputs a code encoded by the first encoding method, decodes by the first decoding method, and outputs a decoded signal;
A spectral parameter calculation unit for inputting the decoded signal and calculating a spectral parameter representing spectral characteristics;
An adaptive codebook unit that calculates a pitch period from the decoded signal and generates an adaptive codebook component based on the pitch period and a past sound source signal;
A noise generator for generating a noise signal;
A coefficient calculating section that converts the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
A gain unit that outputs a sound source signal by adding a gain to at least one of the output signal of the noise generation unit and the output signal of the adaptive codebook unit;
A synthesis filter having a synthesis filter configured using the filter coefficient from the coefficient calculation unit, inputting a sound source signal from the gain unit, passing the synthesis signal through the synthesis filter, and outputting a signal in a band necessary for band conversion And
A sampling frequency conversion circuit that outputs a signal obtained by converting the decoded signal at a predetermined sampling frequency;
An adder for adding and outputting the output signal of the sampling frequency conversion circuit and the output signal of the synthesis filter unit;
A second encoding unit that inputs an output signal of the adder, performs encoding by the second encoding method, and obtains and outputs a second code;
A transcoder characterized by comprising:

In a transcoder for interconverting a code according to the first encoding scheme and a code according to the second encoding scheme,
A first decoding unit that inputs a code encoded by the first encoding method, decodes by the first decoding method, and outputs a decoded signal;
A spectral parameter calculation unit for inputting the decoded signal and calculating a spectral parameter representing spectral characteristics;
An adaptive codebook unit that calculates a pitch period from the decoded signal and generates an adaptive codebook component based on the pitch period and a past sound source signal;
A noise generator for generating a noise signal;
A coefficient calculating section that converts the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
A gain unit that outputs a sound source signal by adding a gain to at least one of the output signal of the noise generation unit and the output signal of the adaptive codebook unit;
A pitch prefilter that performs prefiltering on the sound source signal from the gain unit using the pitch period;
A synthesis filter configured to use the filter coefficient from the coefficient calculation unit, and output a signal in a band necessary for band conversion by passing the output signal of the pitch prefilter through the synthesis filter;
A sampling frequency conversion circuit that outputs a signal obtained by converting the decoded signal at a predetermined sampling frequency;
An adder for adding and outputting the output signal of the sampling frequency conversion circuit and the output signal of the synthesis filter unit;
A second encoding unit that inputs an output signal of the adder, performs encoding by the second encoding method, and obtains and outputs a second code;
A transcoder characterized by comprising:

In a transcoder that mutually converts a code according to the first encoding scheme and a code according to the second encoding scheme,
A first decoding unit that inputs a code encoded by the first encoding method, decodes by the first decoding method, and outputs a decoded signal;
A spectral parameter calculation unit for inputting the decoded signal and calculating a spectral parameter representing spectral characteristics;
A periodic signal generator for calculating a pitch period from the decoded signal and generating a periodic signal using the pitch period;
A noise generator for generating a noise signal;
A coefficient calculating section that converts the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
A gain unit that outputs a sound source signal by adding a gain to at least one of the output signal of the noise generation unit and the output signal of the periodic signal generation unit;
A synthesis filter having a synthesis filter configured using the filter coefficient from the coefficient calculation unit, inputting a sound source signal from the gain unit, passing the synthesis signal through the synthesis filter, and outputting a signal in a band necessary for band conversion And
A sampling frequency conversion circuit that outputs a signal obtained by converting the decoded signal at a predetermined sampling frequency;
An adder for adding and outputting the output signal of the sampling frequency conversion circuit and the output signal of the synthesis filter unit;
A second encoding unit that inputs an output signal of the adder, performs encoding by the second encoding method, and obtains and outputs a second code;
A transcoder characterized by comprising:

In a transcoder for interconverting a code according to the first encoding scheme and a code according to the second encoding scheme,
A first decoding unit that inputs a code encoded by the first encoding method, decodes by the first decoding method, and outputs a decoded signal;
A spectral parameter calculation unit for inputting the decoded signal and calculating a spectral parameter representing spectral characteristics;
A periodic signal generator for calculating a pitch period from the decoded signal and generating a periodic signal using the pitch period;
A noise generator for generating a noise signal;
A coefficient calculating section that converts the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
A gain unit that outputs a sound source signal by adding a gain to at least one of the output signal of the noise generation unit and the output signal of the periodic signal generation unit;
A pitch prefilter that performs prefiltering on the sound source signal from the gain unit using the pitch period;
A synthesis filter configured to use the filter coefficient from the coefficient calculation unit, and output a signal in a band necessary for band conversion by passing the output signal of the pitch prefilter through the synthesis filter;
A sampling frequency conversion circuit that outputs a signal obtained by converting the decoded signal at a predetermined sampling frequency;
An adder for adding and outputting the output signal of the sampling frequency conversion circuit and the output signal of the synthesis filter unit;
A second encoding unit that inputs an output signal of the adder, performs encoding by the second encoding method, and obtains and outputs a second code;
A transcoder characterized by comprising:

The low pass filter which inputs the output signal of the said adaptive codebook part, and passes and outputs the signal of the frequency below a predetermined cut-off frequency is provided. Transcoder.

A post filter configured using a weighting coefficient obtained by weighting the filter coefficient from the count calculation unit;
Regenerate the band-converted signal by passing the output signal of the synthesis filter unit through the post filter,
The adder adds the output signal of the post filter with the output signal of the sampling frequency conversion circuit instead of the output signal of the synthesis filter unit, and outputs the added signal. The transcoder as described in any one of.

The adaptive codebook unit inputs the decoded signal from the first decoding unit and outputs voiced / unvoiced discrimination information and pitch cycle information, and a pitch period from the voiced / unvoiced discrimination circuit, and the synthesis filter unit 4. The transcoder according to claim 2, further comprising an adaptive codebook circuit that receives an input sound source signal.

The periodic signal generator receives the pitch signal from the voiced / unvoiced discrimination circuit that inputs the decoded signal from the first decoding unit and outputs voiced / unvoiced discrimination information and pitch cycle information. 6. The transcoder according to claim 4, further comprising a circuit.

The pitch prefilter receives a pitch period from a voiced / unvoiced discrimination circuit that inputs the decoded signal from the first decoding unit and outputs voiced / unvoiced discrimination information and pitch cycle information. 6. The transcoder according to claim 3, wherein pitch pre-filtering is performed on the sound source signal and then output to the synthesis filter unit.

A voiced / unvoiced discrimination circuit that inputs the decoded signal from the first decoding unit and outputs voiced / unvoiced discrimination information and pitch period information;
A gain adjustment circuit that inputs voiced / unvoiced discrimination information from the voiced / unvoiced discrimination unit and adjusts the gain of a noise signal according to whether voiced or unvoiced;
With
The transcoder according to claim 1, wherein the gain unit includes a gain circuit that receives a gain signal from the gain adjustment circuit and gives a gain to an output signal from the noise generation unit.

A voiced / unvoiced discrimination circuit that inputs the decoded signal from the first decoding unit and outputs voiced / unvoiced discrimination information and pitch period information;
A gain adjustment circuit that inputs voiced / unvoiced discrimination information from the voiced / unvoiced discrimination unit, and adjusts the gain of the adaptive codebook signal and the output signal from the noise generation unit according to whether voiced or unvoiced,
With
The gain unit inputs a gain signal from the gain adjustment circuit, multiplies the output signal of at least one of the adaptive codebook unit and the noise generation unit by a gain, and outputs a gain circuit;
A second adder for adding and outputting two types of signals corresponding to the output signals of the adaptive codebook unit and the noise generation unit output from the gain circuit;
With
4. The transcoder according to claim 2, wherein an output signal of the second adder is supplied to the synthesis filter unit and the adaptive codebook unit.

A voiced / unvoiced discrimination circuit that inputs the decoded signal from the first decoding unit and outputs voiced / unvoiced discrimination information and pitch period information;
A gain adjustment circuit that inputs voiced / unvoiced discrimination information from the voiced / unvoiced discrimination unit, and adjusts the gain of the output signal from the periodic signal generation unit and the noise generation unit according to whether voiced or unvoiced,
With
The gain unit inputs a gain signal from the gain adjustment circuit, multiplies the gain signal by at least one output signal of the periodic signal generation unit and the noise generation unit, and outputs a gain circuit;
A second adder that adds and outputs two types of signals output from the gain circuit and corresponding to the output signals of the periodic signal generator and the noise generator;
With
6. The transcoder according to claim 4, wherein an output signal of the second adder is supplied to the synthesis filter unit.

The synthesis filter unit reproduces a signal whose band is extended,
The sampling frequency conversion circuit receives the decoded signal from the first decoding unit, up-samples to a predetermined sampling frequency, and outputs the sampling signal. A transcoder according to claim 1.

14. The noise generation unit according to claim 1, wherein the average amplitude is normalized to a predetermined level and a band-limited noise signal is generated for a time length equal to a frame length. Transcoder described in 1.

In a transcoder code conversion method for mutually converting a code according to a first encoding method and a code according to a second encoding method,
Decoding a code encoded by the first encoding scheme by a first decoding scheme and outputting a decoded signal;
Calculating and outputting spectral parameters representing spectral characteristics from the decoded signal;
And converting the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
Giving a gain to the output signal from the noise generator;
Passing the output signal given the gain to the synthesis filter configured using the filter coefficient, and outputting a signal in a band necessary for band conversion;
Adding a signal obtained by converting the decoded signal at a predetermined sampling frequency and an output signal of the synthesis filter;
Encoding the addition result by the second encoding method to obtain and output a second code;
A code conversion method comprising:

In a transcoder code conversion method for mutually converting a code according to a first encoding method and a code according to a second encoding method,
Decoding a code encoded by the first encoding scheme by a first decoding scheme and outputting a decoded signal;
Calculating and outputting spectral parameters representing spectral characteristics from the decoded signal;
Calculating a pitch period from the decoded signal, and generating an adaptive codebook component based on the pitch period and a past sound source signal;
And converting the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
Adding a gain to at least one of the noise output from the noise generation unit and the adaptive codebook component and outputting the sound source signal; and
Outputting a signal of a band necessary for band conversion through the sound source signal to a synthesis filter configured using the filter coefficient;
Adding a signal obtained by converting the decoded signal at a predetermined sampling frequency and an output signal of the synthesis filter;
Encoding the addition result by the second encoding method to obtain and output a second code;
A code conversion method comprising:

In a transcoder code conversion method for interconverting a code according to a first encoding scheme and a code according to a second encoding scheme,
Decoding a code encoded by the first encoding scheme by a first decoding scheme and outputting a decoded signal;
Calculating and outputting spectral parameters representing spectral characteristics from the decoded signal;
Calculating a pitch period from the decoded signal, and generating an adaptive codebook component based on the pitch period and a past sound source signal;
And converting the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
Adding a gain to at least one of the noise output from the noise generation unit and the adaptive codebook component and outputting the sound source signal; and
Performing a pitch prefilter on the sound source signal using the pitch period;
Outputting a signal of a band necessary for band conversion through the pitch prefiltered signal to a synthesis filter configured using the filter coefficient;
Adding a signal obtained by converting the decoded signal at a predetermined sampling frequency and an output signal of the synthesis filter;
Encoding the addition result by the second encoding method to obtain and output a second code;
A code conversion method comprising:

In a transcoder code conversion method for mutually converting a code according to a first encoding method and a code according to a second encoding method,
Decoding a code encoded by the first encoding scheme by a first decoding scheme and outputting a decoded signal;
Calculating and outputting spectral parameters representing spectral characteristics from the decoded signal;
Calculating a pitch period from the decoded signal and generating a periodic signal using the pitch period;
And converting the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
Adding a gain to at least one of the noise output from the noise generating unit and the periodic signal and outputting the sound source signal; and
Outputting a signal of a band necessary for band conversion through the sound source signal to a synthesis filter configured using the filter coefficient;
Adding a signal obtained by converting the decoded signal at a predetermined sampling frequency and an output signal of the synthesis filter;
Encoding the addition result by the second encoding method to obtain and output a second code;
A code conversion method comprising:

In a transcoder code conversion method for interconverting a code according to a first encoding scheme and a code according to a second encoding scheme,
Decoding a code encoded by the first encoding scheme by a first decoding scheme and outputting a decoded signal;
Calculating and outputting spectral parameters representing spectral characteristics from the decoded signal;
Calculating a pitch period from the decoded signal and generating a periodic signal using the pitch period;
And converting the filter coefficients in terms of shifting the frequency of the spectrum parameter to a higher frequency,
Adding a gain to at least one of the noise output from the noise generating unit and the periodic signal and outputting the sound source signal; and
Performing a pitch prefilter on the sound source signal using the pitch period;
Outputting a signal of a band necessary for band conversion through the pitch prefiltered signal to a synthesis filter configured using the filter coefficient;
Adding a signal obtained by converting the decoded signal at a predetermined sampling frequency and an output signal of the synthesis filter;
Encoding the addition result by the second encoding method to obtain and output a second code;
A code conversion method comprising:

Regenerating the band-converted signal through the output signal of the synthesis filter unit to a post filter configured using a weighting coefficient obtained by weighting the filter coefficient;
In place of the output signal of the synthesis filter, adding the output signal of the post filter to a signal obtained by converting the decoded signal at a predetermined sampling frequency, and outputting the signal;
21. The code conversion method according to any one of claims 16 to 20, characterized by comprising:

The code conversion method according to claim 17 or 18, further comprising a step of performing low-pass filter processing on the adaptive codebook component.