JP2005004119A

JP2005004119A - Sound signal encoding device and sound signal decoding device

Info

Publication number: JP2005004119A
Application number: JP2003170201A
Authority: JP
Inventors: Takao Yamabe; 孝朗山辺
Original assignee: Victor Company of Japan Ltd
Current assignee: Victor Company of Japan Ltd
Priority date: 2003-06-16
Filing date: 2003-06-16
Publication date: 2005-01-06

Abstract

<P>PROBLEM TO BE SOLVED: To solve problems in which a conventional sound signal encoding device increases in time or processing quantity with the frequency of encoding of a difference signal though circuit constitution concerning encoding and decoding can be used in common and can not precisely decode the encoded difference signal. <P>SOLUTION: A difference calculator 5 finds a frequency area difference signal which is the value of the difference between frequency samples before and after quantization by a quantizer 3. The frequency sample before quantization is sent from a frequency conversion part 1 and the frequency sample after reverse quantization is sent from a reverse quantizer 4 to the difference calculator 5 respectively. The frequency area difference signal is encoded by a difference information encoding part 6 into a frequency area difference signal encoded signal, which is transmitted together with a sound signal encoded signal of a 1st layer from the quantizer 3. A sound signal which compares with a source sound signal with extremely high precision is obtained by making corrections in a frequency area before frequency reverse conversion by using the sound signal encoded signal and frequency area difference signal in decoding. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

【０００１】
【発明の属する技術分野】
本発明は音響信号符号化装置及び音響信号復号化装置に係り、特に音響信号を複数の周波数領域に分割し、各分割周波数領域毎の重み付けに基づき決定された量子化精度に従って量子化を行い情報圧縮した音響信号符号化信号を生成する音響信号符号化装置及び音響信号復号化装置に関する。
【０００２】
【従来の技術】
音響信号を周波数領域に変換した後、周波数領域にて情報圧縮を行う符号化装置または符号化方法には、ビデオＣＤ（ｃｏｍｐａｃｔｄｉｓｋ）等で用いられるＭＰＥＧ−１ＡＵＤＩＯＬａｙｅｒ２、インターネット等のデータ配信で利用されているＭＰＥＧ−１ＡＵＤＩＯＬａｙｅｒ３（略称ＭＰ３）、ＢＳディジタル放送や地上波ディジタル放送で採用されているＭＰＥＧ−２ＡＵＤＩＯＮＢＣ（略称ＡＡＣ）、ＤＶＤ（ｄｉｇｉｔａｌｖｅｒｓａｔｉｌｅｄｉｓｋ）の音声フォーマットであるドルビーディジタル（ＤｏｌｂｙＤｉｇｉｔａｌ）、ＭＤ（ｍｉｎｉｄｉｓｋ）で採用されている圧縮方式ＡＴＲＡＣ（ａｄａｐｔｉｖｅｔｒａｎｓｆｏｒｍａｃｏｕｓｔｉｃｃｏｄｉｎｇ）またはＡＴＲＡＣ３などがある。
【０００３】
これらは時間領域の音響信号を周波数領域に変換し、特定の周波数帯に偏りを持つ音響信号の特徴と、人間の聴覚の特性を考慮した聴覚の感度に応じた周波数帯毎の重み付けに従い、聴感的に重要とされない周波数帯の情報を削減または減少することによって情報圧縮を行っている。
【０００４】
これら情報圧縮を行った音響信号符号化信号を復号した再生信号には、聴感特性上、劣化は分かり難いものの、元の音響信号と比較した場合、相当量を有する量子化ノイズが含まれており、ＣＤ等の非圧縮のリニアＰＣＭで符号化された音響信号符号化信号を復号した再生信号に対し明らかに劣化している。
【０００５】
ところで、近年、オーディオの嗜好形態は多様化しており、手軽に持ち運びが可能な携帯再生装置や、伝送路の帯域や蓄積媒体の容量の不足を補うために便利な高能率符号化技術を用いた上記のような圧縮オーディオが重宝される一方、ＣＤも含めＤＶＤ−ＡｕｄｉｏやＳＡＣＤ（ＳｕｐｅｒＡｕｄｉｏＣＤ）に代表される高品位な非圧縮オーディオに対しても注目が集まっている。
【０００６】
そのような要求に応える方法として、圧縮オーディオの品質をより原音に近づける、または原音と変わらない可逆符号化を提供する音響信号符号化装置及び復号化システムが従来より知られている（例えば、特許文献１参照）。この特許文献１記載の従来の音響信号符号化装置は、圧縮された音響信号を復号化し再び時間領域の信号に戻した後、原音響信号と復号化後音響信号との差分信号を再圧縮し、一次符号化の際に欠落した音響信号の信号成分を、差分信号である二次符号化信号若しくは高次符号化信号によって補うものである。
【０００７】
この従来装置の特徴として、既に存在する符号化装置及び復号装置を再利用する事で、安価な構成で圧縮された音響信号の高音質化を達成できることにあり、差分信号の情報量を増加する事によって可逆符号化をも実現することができる点にある。
【０００８】
【特許文献１】
特開平８−４６５１７号公報
【０００９】
【発明が解決しようとする課題】
しかしながら、上記の特許文献１記載の従来装置は、入力音響信号を符号化回路により符号化処理して得た第１の符号化信号を、復号化回路により復号化処理し、その復号音響信号と上記の入力音響信号との差分信号を差分算出回路により算出し、その差分信号を符号化回路により更に符号化処理して得た第２の符号化信号を、上記の第１の符号化信号と共に出力する構成であり、符号化及び復号化に関わる回路構成を共用することが可能であるが、その時間または処理量は約２倍、３倍と差分信号を符号化する回数に従い増加する。
【００１０】
また、音響信号を符号化する符号化回路は音響特性（聴感特性）を利用して情報圧縮するため、必ずしも音響信号とはいえない差分信号に適用した場合、差分信号を補正しようとする従来装置の目的を達成するには適当ではない。
【００１１】
更に、上記の一般的な音響信号符号化方式は、人間の聴感特性の一つである、周波数帯によって決まる、あるレベル以下の信号は聴こえないとされる最小可聴限特性を情報削減の指標に用いている。この最小可聴限特性は図１０に示すように、低域周波数帯や高域周波数帯の音響信号は知覚され難いことが知られており、これら周波数帯の信号は常に情報削減の対象となる部分であり、複数回差分信号の符号化を繰り返しても、符号化された差分信号を精度良く復元する事は非常に困難である。
【００１２】
加えて、音響信号符号化信号には補助情報と呼ばれる、音響信号への復元に必要な情報が含まれているが、このような情報を繰り返し伝送する事は、符号化情報量の増大をもたらし、伝送情報量によって得られる信号品質の向上効果は、情報量の割には希薄であることが推測される。
【００１３】
本発明は以上の点に鑑みなされたもので、音響信号符号化信号の音質向上を図るべく、符号化過程で生じる周波数スペクトルの量子化前後における差分値を求め、周波数領域差分情報として、符号化信号と多重化または別々に伝送し、復号時には音響信号符号化信号と周波数領域差分信号を用いて、周波数逆変換前に周波数領域にて補正を行う事で、極めて高い精度で原音響信号に匹敵する音響信号を、または蓄積媒体若しくは伝送路の状況に応じたより高い品質の音響信号を得る音響信号符号化装置及び音響信号復号化装置を提供することを目的とする。
【００１４】
【課題を解決するための手段】
上記の目的を達成するため、第１の発明の音響信号符号化装置は、音響信号の複数の周波数領域毎の周波数サンプルの量子化を行って得られた第一階層を成す音響信号符号化信号と、自身よりも上位階層に含まれる量子化誤差を補正するための信号を符号化して得た第二階層以下の階層を成す符号化信号とを出力する音響信号符号化装置であって、供給された音響信号を複数の周波数領域に分割し、聴感特性を用いた各周波数領域毎の重み付けに基づき決定された量子化精度に従って、音響信号の各周波数領域毎の周波数サンプルの量子化を行うことにより、情報圧縮した第一階層を成す音響信号符号化信号を生成する符号化器と、符号化器において生成された音響信号符号化信号を逆量子化して得られた、各周波数領域毎の周波数サンプルの逆量子化値と、符号化器に供給された量子化前の音響信号の各周波数領域毎の周波数サンプルとの差分値を符号化して、第二階層以下のの符号化信号である周波数領域差分信号符号化信号を生成する周波数領域差分信号符号化器とを有する構成としたものである。
【００１５】
また、上記の目的を達成するため、第２の発明の音響信号符号化装置は、第１の発明の符号化器によって生成された音響信号符号化信号と、周波数領域差分信号符号化器によって生成された周波数領域差分信号符号化信号とを、時系列を揃えた状態で多重化して出力する第１の機能と、音響信号符号化信号との時系列を一致させるための時系列情報を周波数領域差分信号符号化信号に付加して、時系列情報を付加した周波数領域差分信号符号化信号と音響信号符号化信号とを各別に出力する第２の機能との少なくとも一方の機能を具備する多重化装置を有することを特徴とする。
【００１６】
第１及び第２の発明では、従来の音響信号符号化装置に含まれていた量子化誤差を周波数領域の差分信号として符号化した周波数領域差分信号符号化信号を第二階層以下の符号化信号として、第一階層の音響信号符号化信号と組み合わせることにより、階層構造をなす符号化信号を生成することができる。
【００１７】
また、上記の目的を達成するため、第３の発明の音響信号復号化装置は、音響信号を複数の周波数領域に分割し、聴感特性を用いた各周波数領域毎の重み付けに基づき決定された量子化精度に従って、音響信号の各周波数領域毎の周波数サンプルの量子化を行うことにより生成された第一階層を成す音響信号符号化信号と、音響信号符号化信号を逆量子化して得られた、各周波数領域毎の周波数サンプルの逆量子化値と、量子化前の音響信号の各周波数領域毎の周波数サンプルとの差分値を符号化することにより生成された、第二階層以下の周波数領域差分信号符号化信号とが供給され、音響信号符号化信号と周波数領域差分信号符号化信号とを時系列を一致させた状態で各別に出力する分離手段と、分離手段により出力された音響信号符号化信号を逆量子化して、周波数サンプル逆量子化値を取得する逆量子化手段と、分離手段により出力された周波数領域差分信号符号化信号を復号して周波数領域差分信号を得る周波数領域差分信号復号手段と、周波数サンプル逆量子化値と周波数領域差分信号を加算する周波数サンプル加算手段と、周波数サンプル加算手段から出力された周波数サンプルを時間信号に変換する周波数逆変換手段とを有する構成としたものである。
【００１８】
第３の発明では、従来の音響信号符号化装置に含まれていた量子化誤差を周波数領域の差分信号として符号化した周波数領域差分信号符号化信号を、第一階層の音響信号符号化信号と組み合わせた階層構造をなす符号化信号を復号化するようにしたため、復号過程で、第一階層である音響信号符号化信号中の周波数スペクトルをより正確な信号に補正し、原信号に近い、または同等の再生音を提供する事ができる。
【００１９】
【発明の実施の形態】
次に、本発明の一実施の形態について図面と共に説明する。図１は本発明になる音響信号符号化装置の一実施の形態のブロック図を示す。同図において、符号化されるべきディジタル音響信号は周波数変換部１に送られ、ここで時間領域の信号から周波数領域の信号へと変換される。ここで用いる周波数変換法は、第一階層を形成する音響圧縮方式に従う。
【００２０】
例えば、音響圧縮方式がＭＰＥＧ−１Ａｕｄｉｏであれば周波数変換部１はポリフェイズフィルタで構成され、音響圧縮方式がＭＰＥＧ−１Ｌａｙｅｒ３であれば、周波数変換部１はポリフェィズフィルタとＭＤＣＴフィルタを組み合わせたハイブリッドフィルタで構成され、音響圧縮方式がＡＡＣ、ＡＴＲＡＣ、ドルビーディジタル等であれば、周波数変換部１はＭＤＣＴフィルタで構成される。また、周波数変換長も各方式で異なるが、どの方式であっても符号化処理は周波数変換部１で採用した変換長を単位として以降の処理が実行される。
【００２１】
また、符号化されるべきディジタル音響信号は周波数変換部１に供給されると共に、聴覚心理モデル２へも供給される。この聴覚心理モデル２は人間の聴感特性を考慮し周波数帯域毎に重み付けを行い、符号化方式によって定められたサブバンド（周波数帯）毎に量子化の精度を決定する部分である。
【００２２】
周波数変換部１を経て周波数領域上の信号に変換された周波数サンプルは、量子化器３に供給され、ここで聴覚心理モデル２にて算定された複数の周波数領域に分割されて各周波数領域毎の重み付けに基づき決定された量子化精度に従って、各周波数領域毎に量子化されて、量子化後周波数サンプルとされる。この時点で聴覚心理モデル２において聴感上検知されないと判断された量子化誤差が量子化後周波数サンプルに含まれる事になる。
【００２３】
量子化器３から取り出された量子化後周波数サンプルは逆量子化器４に供給され、ここで逆量子化されて、元の周波数スペクトルと同様の振幅値を持つ周波数信号へと復元される。ここで用いる周波数逆変換部は後述の復号化装置に含まれる周波数逆変換部に等しい。逆量子化で必要な量子化精度情報などは、聴覚心理モデル２から供給される。
【００２４】
続いて、差分算出器５において量子化前後における周波数サンプル間の差分値である周波数領域差分信号が求められる。ここで、量子化前周波数サンプルは周波数変換部１から、逆量子化後の周波数サンプルは逆量子化器４からそれぞれ差分算出器５に送られてくる。
【００２５】
次に、差分算出器５において算出された周波数領域差分信号は、差分情報符号化部６において符号化される。ここで差分情報符号化部６は、周波数領域差分信号の全ビットを同一部分として符号化してもよいし、周波数領域差分信号のビット列を揃え、それを情報の重要度に応じて、１ビットないしは複数ビットを単位として分割し、それぞれ別のグループとして（別の階層として）符号化してもよい。１または複数の部分にまとめられた周波数領域差分信号は、差分値の振幅や周波数帯毎の区分値などの補助情報と共に差分情報符号化部６において符号化され、周波数領域差分信号符号化信号とされて多重化／符号化情報形成部７へ供給される。差分情報符号化部６については後に更に詳しく説明する。
【００２６】
多重化／符号化情報形成部７は少なくとも第一階層をなす単体で復元可能な音響信号符号化信号を、聴覚心理モデル２からの補助情報と量子化器３からの量子化後周波数サンプルを用いて形成する。このまま一つの符号化信号として伝送してもよいし、更に周波数領域差分信号符号化信号と共に多重化し伝送してもよい。多重化して伝送する場合は、例えば図２で示したような各階層の符号化信号が時系列的に合成されたビット列として伝送する。
【００２７】
図２において、第一階層に含まれる周波数サンプルと周波数領域の差分情報として付加された第二階層以下の周波数サンプルは、時系列が一致した状態で符号化及び多重化されるため、復号化器ではこれら両者の情報を復号化することによって、直ちに品質の高い復号後信号を得る事ができる。なお、第二階層以下の周波数領域差分信号符号化信号の情報は、第一階層符号化部において欠落した情報であり、原信号と第一階層を符号化する際に発生した周波数領域上の差分信号をビットの重要度や必要であれば時間情報、階層番号などの補助情報と周波数領域の差分信号そのものを符号化した情報である。なお、図２において、「ＳＹＮＣ」は同期情報である。
【００２８】
また、単体で復元可能な第一階層の音響信号符号化信号とは別に、周波数領域差分情報を符号化し伝送する場合は、第一階層とのフレーム単位での時系列一致を図るため、識別信号を付加して伝送する。フレーム毎に異なる識別信号であることが望ましいため、例えば、図３に示すように、第一階層の音響信号符号化信号が持つフレーム番号を第二階層以下の周波数領域差分信号符号化信号の識別信号として付加する方式がとれる。
【００２９】
なお、第一階層の音響信号符号化信号の先頭フレームと第二階層以下の周波数領域差分信号符号化信号の先頭のフレーム番号が一致し、かつ、周波数領域差分信号符号化信号が常に存在する場合は、時系列一致を示す識別信号を付加する必要はない。
【００３０】
更に第一階層をなす音響信号符号化信号には、エラー検出コードが含まれている場合が多い。このエラー検出コードは、同一のコードが出現する確率は、音響信号が無音の場合を除いて極めて低いため、時系列の一致を示す識別信号として利用する事ができる。このような方式をとる場合には、図４に示すように、第二階層以下の符号化信号には第一階層で作成されたエラー検出コードを付加すればよい。図４は一般的な１６ビットのＣＲＣエラー検出コードを識別信号とした例であり、図中の識別信号は１６進の数値で表示してある。
【００３１】
また、第一階層の音響信号符号化信号においてフレーム毎の変化が期待される部分に、サンプル符号化部分が挙げられる。図５に示すように、ある特定の周波数帯におけるサンプル群（ビット列）、またはフレーム内の特定可能なサンプル符号化部のビット列を時系列一致用の識別信号として採用し、第二階層以下の周波数領域差分信号符号化信号に付加する事で、周波数サンプルの補正を行う事ができる。
【００３２】
上記の特定部分は、例えばフレーム先頭から決められたジャンプ先から数ビット、または特定の周波数帯サンプルの位置を占めるビット列数ビットを用いる。
図５は第一階層の量子化後の周波数サンプルの部分に位置するフレーム先端から１００ビット目以降１６ビットを時間軸一致を示す識別番号に用いた例である。
【００３３】
時間一致を示す識別信号はフレームが時間の流れと共に更新していくため、常にフレーム内に付加される必要はなく、ある間隔を置いて挿入する事も可能である。また、無音区間や第一階層の音響信号符号化信号の品質が良好の場合は、第二階層以下の周波数領域差分信号符号化信号を伝送せずに済ます事もできる。このような状況に対応するため、伝送しない間隔の前後のフレームに時系列一致を示す識別信号を付加しておけば、第一階層の音響信号符号化信号との整合性が維持される。
【００３４】
以上のような手段を用いて、図１の多重化／符号化情報形成部７では一又は二以上の符号化信号を形成し出力する。
【００３５】
次に、図１に示した音響信号符号化装置に含まれる差分情報符号化部６について詳しく説明する。音響信号を高精度に復元する場合、すなわち原音に匹敵する演算精度を持たせた可逆符号化信号を作成するには、差分算出器５で求められた周波数領域差分信号を全て符号化し伝送すればよい。
【００３６】
しかしながら、段階を追って階層的に復号化後の音響信号の品質の向上を図るには、ビットの重要度別に複数の階層に分割し伝送する方が適当である。また、ＣＤ品質のみならず、量子化ビット長が１６ビットを超えるＤＶＤＡｕｄｉｏや一部のＤＶＤＶｉｄｅｏ等の品質を持つ音響信号を符号化する場合は、特に複数の階層に分割することで、様々な形態の蓄積メディアや伝送路への適応性を持たせる事ができる。
【００３７】
差分算出器５において算出された周波数領域差分信号は、符号化されなかった音楽的な成分が残されているものの、基本的には量子化雑音であり、ランダムノイズ（白色雑音）として扱う事ができる。但し、雑音のパワーには周波数帯域毎のばらつきがある。これは第一階層の音響信号符号化処理にて行われた周波数帯域毎に個別に設定された量子化精度の違いによるものである。
【００３８】
図６は周波数領域差分信号の一例の周波数対エネルギー変化量（パワー）特性図を示す。この特性図は音響信号符号化処理における周波数変換後の信号と量子化・逆量子化を経た後の周波数信号との差分のパワーを周波数軸上に図示したものである。図６に示すように量子化雑音であるこの差分信号は、周波数帯域毎に異なり一様ではない。
【００３９】
周波数領域差分信号を符号化するには、この周波数帯域毎に異なる振幅情報を補助情報として、周波数領域差分信号と共に符号化する事によって符号化効率を向上する事が可能である。また、図７に示すように、周波数領域差分信号を所定の周波数帯域毎に区切り、その各周波数帯域内（バンド１〜バンドｎ）の最大振幅値を持つ信号を再量子化するに足りるビット長でもって、帯域内差分信号を符号化してもよい。図７の例では周波数領域差分信号は全て正の数として扱っているが、別に正の数に限定したものではない。負の信号には符号ビットと絶対値で表現した振幅値で対応する事ができる。
【００４０】
また、別の方法として、フレーム内の周波数領域差分信号最大振幅値を基準に、フレーム内再量子化値を求め、全差分信号に渡りビットの重要度別にグループ化し符号化する方法もある。このときグループ化されるビット列は１ビット単位でもよいし、複数ビットをグループ化して符号化してもよい。この場合、上位ビットは多分に冗長性を持つため、ハフマン符号化や算術符号化のようなエントロピー符号化を利用すると効果的である。
【００４１】
図８の例では１ビットを単位にグループ化し符号化した場合を示している。負の差分信号は符号ビットを除きビット反転する事でビットが「０」となる確率が増え、エントロピー符号化の効果が期待される。エントロピー符号化器によりビット列グループ別に符号化することによって容易に階層構造を持つ事ができる。
【００４２】
ここで、周波数領域差分信号を符号化する際どちらか一方の符号化方式に限定されるものではなく、例えば上位ビット群を前者の周波数帯による再量子化情報を含むバンド別符号化方式を採用し、下位ビット群をフレーム全体でビットの重要度別にグループ化し符号化する方式を採用してもよく、組み合わせて使う事ができる。
【００４３】
次に、本発明の音響信号復号化装置について図９と共に説明する。図９は本発明になる音響信号復号化装置の一実施の形態のブロック図を示す。同図において、ＤｅＭＵＸ／符号化情報分類部８は、図１の音響信号符号化装置から出力された１又は複数の符号化信号を入力として受け、分離化して時系列が一致する音響信号符号化信号と周波数領域差分情報符号化信号とに分類する。
【００４４】
すなわち、ＤｅＭＵＸ／符号化情報分類部８は、単独で復元可能な音響信号符号化信号の第一階層部と多重化された状態で第二階層以下の周波数領域差分信号符号化信号が入力されるとき、直ちに符号化信号を分離し、第一階層に含まれる量子化された音響信号符号化信号は逆量子化器９へ供給すると共に、周波数領域差分情報符号化信号は差分情報復号化部１０へ供給する。
【００４５】
一方、ＤｅＭＵＸ／符号化情報分類部８は、階層別に別々に符号化信号が入力されるときは、時系列一致を図るため、前述の時系列一致を示す識別信号を検索し、第一階層に第二階層以下の符号化信号を揃えなければならないため、フレーム同期機能を有しており、時系列が一致した状態で分類された第一階層の符号化信号はそれぞれの音響信号符号化方式に従った逆量子化器９に供給して逆量子化させて、本来の振幅値を持つ周波数サンプルへと変換させる。
【００４６】
一方、第二階層以下の周波数領域差分信号符号化信号は差分情報復号化部１０に供給され、同様に本来の振幅値を持つ周波数領域の差分信号として復元される。第二階層以下の周波数領域差分信号符号化信号が複数存在する場合は複数回処理を繰り返すか多段構成をとる事で周波数領域差分信号を取得できる。ここで、音響信号符号化器における差分情報符号化部６でエントロピー符号化が施された場合、この差分情報復号化部１０でエントロピー符号化信号を復号する。
【００４７】
逆量子化器９で音響信号符号化信号を逆量子化して得られた複数の周波数領域の周波数サンプル（周波数スペクトル逆量子化値）は、周波数サンプル加算部１１に供給され、ここで差分情報復号化部１０で復号化された周波数領域差分情報と加算され、量子化誤差を多分に含んでいた第一階層の音響信号符号化信号の周波数サンプルをより正確な周波数サンプルへと補正される。
【００４８】
最後に、周波数サンプル加算部１１から取り出された周波数サンプルは、周波数逆変換部１２に供給され、ここで周波数領域の信号が時間領域の信号に変換されて音響信号として出力される。ここで用いる周波数逆変換法は第一階層の音響符号化方式で用いる周波数変換法と同一のものである。
【００４９】
以上のように、本実施の形態では周波数帯域の量子化誤差を補正する符号化信号を生成する周波数領域差分信号符号化器を備えることで、従来の音響符号化方式にて生成された符号化信号に比較して、より良い品質の再生音を提供する事が可能であり、また、蓄積媒体の容量や伝送の帯域の状況に応じて一定の品質を維持しつつ、柔軟な情報伝達を行う事ができる。
【００５０】
なお、本発明は上記の実施の形態に限定されるものではなく、音響信号符号化装置には、複数ビットを有する周波数領域差分信号を符号化する際のビットの並びを、振幅値を示す通常のビット列として符号化するか、ビット重要度を揃えた状態で複数の差分サンプルのある１ビットを用いたビット列として符号化するかを選択するビット列生成部及びビット列識別情報を有するようにしてもよい。
【００５１】
【発明の効果】
以上説明したように、本発明の音響信号符号化装置によれば、従来の音響信号符号化装置に含まれていた量子化誤差を周波数領域の差分信号として符号化した周波数領域差分信号符号化信号を第二階層以下の符号化信号として、第一階層の音響信号符号化信号と組み合わせることにより、階層構造をなす符号化信号を生成するようにしたため、狭帯域伝送路や低容量蓄積メディアから広帯域、大容量な媒体にも対応でき、柔軟な情報伝達を実現する事ができる。
【００５２】
また、本発明の音響信号復号化装置によれば、従来の音響信号符号化装置に含まれていた量子化誤差を周波数領域の差分信号として符号化した周波数領域差分信号符号化信号を、音響信号符号化信号と組み合わせた階層構造をなす符号化信号を復号化するようにしたため、復号過程で、第一階層である音響信号符号化信号中の周波数スペクトルをより正確な信号に補正し、原信号に近い、または同等の再生音を提供する事ができる。
【図面の簡単な説明】
【図１】本発明になる音響信号符号化装置の一実施の形態のブロック図である。
【図２】本発明により生成される音響信号符号化信号のビット列の一例を示す図である。
【図３】第一階層のフレーム番号を用いた時系列一致用識別信号の付加方法を示す図である。
【図４】第一階層のエラー検出コードを用いた時系列一致用識別信号の付加方法を示す図である。
【図５】第一階層の特定のサンプル領域部分を用いた時系列一致用識別信号の付加方法を示す図である。
【図６】周波数領域差分信号の一例のパワー対周波数特性図である。
【図７】周波数領域差分信号を帯域毎にグループ化して符号化した例を示す図である。
【図８】周波数領域差分信号をビットの重要度別にグループ化して符号化した例を示す図である。
【図９】本発明になる音響信号復号化装置の一実施の形態のブロック図である。
【図１０】人間の聴感特性の一種である最小可聴限特性を示す図である。
【符号の説明】
１周波数変換部
２聴覚心理モデル
３量子化器
４、９逆量子化器
５差分算出器
６差分情報符号化部
７多重化／符号化情報形成部
８ＤｅＭＵＸ／符号化情報分類部
１０差分情報復号化器
１１周波数サンプル加算部
１２周波数逆変換部[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an acoustic signal encoding device and an acoustic signal decoding device, and in particular, divides an acoustic signal into a plurality of frequency regions and performs quantization according to quantization accuracy determined based on weighting for each divided frequency region. The present invention relates to an audio signal encoding device and an audio signal decoding device that generate a compressed audio signal encoded signal.
[0002]
[Prior art]
An encoding apparatus or encoding method for compressing information in the frequency domain after converting the acoustic signal into the frequency domain includes data distribution such as MPEG-1 AUDIO Layer 2 used in a video CD (compact disk), the Internet, etc. MPEG-1 AUDIO Layer 3 (abbreviated as MP3), MPEG-2 AUDIO NBC (abbreviated as AAC) and DVD (digital versatile disk) used in BS digital broadcasting and terrestrial digital broadcasting. There are compression schemes ATRAC (adaptive transform acoustic coding) or ATRAC3 adopted in Dolby Digital and MD (mini disk).
[0003]
They convert time domain acoustic signals to frequency domain, and perceptually according to the characteristics of acoustic signals that are biased in a specific frequency band and the weighting of each frequency band according to the auditory sensitivity that takes into account human auditory characteristics. Information compression is performed by reducing or reducing information in frequency bands that are not important.
[0004]
The reproduced signal obtained by decoding the information-encoded acoustic signal encoded signal does not easily deteriorate due to auditory characteristics, but contains a considerable amount of quantization noise when compared with the original acoustic signal. The reproduced signal obtained by decoding the encoded audio signal encoded by the uncompressed linear PCM such as CD is clearly deteriorated.
[0005]
By the way, in recent years, audio preferences have been diversified, and portable playback devices that can be easily carried, and high-efficiency coding technology that is convenient to make up for the shortage of transmission path bandwidth and storage medium capacity have been used. While the above compressed audio is useful, attention is also focused on high-quality uncompressed audio represented by DVD-Audio and SACD (Super Audio CD) including CD.
[0006]
As methods for meeting such demands, acoustic signal encoding apparatuses and decoding systems that provide lossless encoding that brings the quality of compressed audio closer to that of the original sound or that does not differ from the original sound have been known (for example, patents). Reference 1). The conventional acoustic signal encoding device described in Patent Document 1 decodes a compressed acoustic signal and returns it to a time domain signal again, and then recompresses the difference signal between the original acoustic signal and the decoded acoustic signal. The signal component of the acoustic signal that is lost during the primary encoding is supplemented by a secondary encoded signal or a high-order encoded signal that is a differential signal.
[0007]
As a feature of this conventional apparatus, by reusing an existing encoding apparatus and decoding apparatus, it is possible to achieve high sound quality of the compressed acoustic signal with an inexpensive configuration, and increase the information amount of the differential signal. Therefore, lossless encoding can also be realized.
[0008]
[Patent Document 1]
JP-A-8-46517
[0009]
[Problems to be solved by the invention]
However, the conventional apparatus described in Patent Document 1 described above decodes the first encoded signal obtained by encoding the input acoustic signal by the encoding circuit using the decoding circuit, A difference signal from the input acoustic signal is calculated by a difference calculation circuit, and the second encoded signal obtained by further encoding the difference signal by the encoding circuit is combined with the first encoded signal. Although it is a configuration for outputting, it is possible to share a circuit configuration related to encoding and decoding, but the time or the processing amount increases approximately twice or three times according to the number of times the difference signal is encoded.
[0010]
In addition, since an encoding circuit that encodes an acoustic signal compresses information using acoustic characteristics (audibility characteristics), when applied to a differential signal that is not necessarily an acoustic signal, a conventional apparatus that corrects the differential signal It is not appropriate to achieve this goal.
[0011]
Furthermore, the above-described general audio signal encoding method uses the minimum audible limit characteristic, which is one of human audibility characteristics, which is determined by the frequency band and cannot be heard as a signal below a certain level, as an indicator of information reduction. Used. As shown in FIG. 10, it is known that the minimum audible limit characteristic is that it is difficult to perceive an acoustic signal in a low frequency band or a high frequency band, and the signal in these frequency bands is always a target of information reduction. Even if the encoding of the differential signal is repeated a plurality of times, it is very difficult to accurately restore the encoded differential signal.
[0012]
In addition, the audio signal encoded signal includes information necessary for restoration to the audio signal called auxiliary information, but repeatedly transmitting such information leads to an increase in the amount of encoded information. It is estimated that the signal quality improvement effect obtained by the amount of transmission information is sparse for the amount of information.
[0013]
The present invention has been made in view of the above points, and in order to improve the sound quality of an audio signal encoded signal, a difference value before and after quantization of a frequency spectrum generated in the encoding process is obtained and encoded as frequency domain difference information. The signal is multiplexed or transmitted separately, and at the time of decoding, the acoustic signal encoded signal and the frequency domain difference signal are used to perform correction in the frequency domain before frequency inversion, which is comparable to the original acoustic signal with extremely high accuracy. It is an object of the present invention to provide an acoustic signal encoding device and an acoustic signal decoding device that obtain an acoustic signal to be transmitted or a higher quality acoustic signal in accordance with the state of a storage medium or a transmission path.
[0014]
[Means for Solving the Problems]
In order to achieve the above object, an audio signal encoding device according to a first aspect of the present invention is an audio signal encoded signal forming a first layer obtained by quantizing frequency samples for a plurality of frequency regions of an audio signal. And an audio signal encoding device that outputs an encoded signal that forms a layer below the second layer obtained by encoding a signal for correcting a quantization error included in a higher layer than itself, Dividing the acoustic signal into a plurality of frequency domains and quantizing the frequency samples for each frequency domain of the acoustic signal according to the quantization accuracy determined based on the weighting of each frequency domain using auditory characteristics By the above, an encoder that generates an audio signal encoded signal that forms the first layer of information compression, and a frequency for each frequency domain obtained by dequantizing the audio signal encoded signal generated by the encoder sample The difference between the inverse quantized value and the frequency sample for each frequency domain of the unquantized acoustic signal supplied to the encoder is encoded, and the frequency domain difference that is the encoded signal in the second and lower layers A frequency domain differential signal encoder that generates a signal encoded signal is provided.
[0015]
In order to achieve the above object, the acoustic signal encoding device of the second invention is generated by the acoustic signal encoded signal generated by the encoder of the first invention and the frequency domain differential signal encoder. The time domain information for matching the time series of the first function to multiplex and output the frequency domain difference signal encoded signal with the time series aligned and the time series of the acoustic signal encoded signal to the frequency domain Multiplexing having at least one function of a frequency domain difference signal encoded signal to which time series information is added in addition to the differential signal encoded signal and a second function for separately outputting the acoustic signal encoded signal It has the apparatus.
[0016]
In the first and second inventions, the frequency domain differential signal encoded signal obtained by encoding the quantization error included in the conventional acoustic signal encoding apparatus as the frequency domain differential signal is encoded signals in the second and lower layers. As described above, by combining with the first-layer acoustic signal encoded signal, an encoded signal having a hierarchical structure can be generated.
[0017]
In order to achieve the above object, an acoustic signal decoding apparatus according to a third aspect of the present invention divides an acoustic signal into a plurality of frequency domains, and determines the quantum determined based on weighting for each frequency domain using auditory characteristics. According to the conversion accuracy, obtained by dequantizing the acoustic signal encoded signal and the acoustic signal encoded signal forming the first layer generated by quantizing the frequency sample for each frequency region of the acoustic signal, Frequency domain difference of the second and lower layers generated by encoding the difference value between the inverse quantized value of the frequency sample for each frequency domain and the frequency sample for each frequency domain of the sound signal before quantization. A signal encoding signal and a separation unit that outputs the acoustic signal encoded signal and the frequency domain difference signal encoded signal separately in a time-sequential state; and the acoustic signal code output by the separation unit Inverse quantization means for dequantizing a signal to obtain a frequency sample inverse quantization value, and frequency domain difference signal decoding for decoding a frequency domain difference signal encoded signal output by a separation means to obtain a frequency domain difference signal And a frequency sample addition means for adding the frequency sample inverse quantization value and the frequency domain difference signal, and a frequency inverse conversion means for converting the frequency sample output from the frequency sample addition means into a time signal. It is.
[0018]
In the third invention, the frequency domain differential signal encoded signal obtained by encoding the quantization error included in the conventional acoustic signal encoding apparatus as the frequency domain differential signal is converted into the first layer acoustic signal encoded signal. Since the encoded signal having the combined hierarchical structure is decoded, in the decoding process, the frequency spectrum in the audio signal encoded signal that is the first layer is corrected to a more accurate signal, close to the original signal, or Equivalent playback sound can be provided.
[0019]
DETAILED DESCRIPTION OF THE INVENTION
Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a block diagram of an embodiment of an acoustic signal encoding apparatus according to the present invention. In the figure, a digital sound signal to be encoded is sent to a frequency converter 1 where it is converted from a time domain signal to a frequency domain signal. The frequency conversion method used here follows the acoustic compression method that forms the first layer.
[0020]
For example, if the acoustic compression method is MPEG-1 Audio, the frequency conversion unit 1 is composed of a polyphase filter, and if the acoustic compression method is MPEG-1 Layer 3, the frequency conversion unit 1 is a polyphase filter and an MDCT filter. If the acoustic compression method is AAC, ATRAC, Dolby Digital, or the like, the frequency converter 1 is configured with an MDCT filter. Further, although the frequency conversion length is different for each method, the encoding processing is performed in units of the conversion length adopted in the frequency conversion unit 1 in any method.
[0021]
Further, the digital acoustic signal to be encoded is supplied to the frequency conversion unit 1 and also to the psychoacoustic model 2. This auditory psychological model 2 is a part that weights each frequency band in consideration of human auditory characteristics and determines the accuracy of quantization for each subband (frequency band) defined by the encoding method.
[0022]
The frequency sample converted into the signal on the frequency domain through the frequency conversion unit 1 is supplied to the quantizer 3, where it is divided into a plurality of frequency domains calculated by the psychoacoustic model 2, and each frequency domain is divided. In accordance with the quantization accuracy determined on the basis of the weighting, the frequency is quantized for each frequency region to obtain a quantized frequency sample. At this time, the quantization error determined not to be detected in the auditory psychology model 2 is included in the quantized frequency sample.
[0023]
The quantized frequency sample taken out from the quantizer 3 is supplied to the inverse quantizer 4, where it is inversely quantized and restored to a frequency signal having the same amplitude value as the original frequency spectrum. The frequency inverse transform unit used here is the same as the frequency inverse transform unit included in the decoding apparatus described later. Quantization accuracy information and the like necessary for inverse quantization are supplied from the auditory psychological model 2.
[0024]
Subsequently, the difference calculator 5 obtains a frequency domain difference signal that is a difference value between frequency samples before and after quantization. Here, the frequency sample before quantization is sent from the frequency conversion unit 1, and the frequency sample after inverse quantization is sent from the inverse quantizer 4 to the difference calculator 5.
[0025]
Next, the frequency domain difference signal calculated by the difference calculator 5 is encoded by the difference information encoding unit 6. Here, the difference information encoding unit 6 may encode all the bits of the frequency domain difference signal as the same part, or arranges a bit string of the frequency domain difference signal and changes the bit string according to the importance of the information. A plurality of bits may be divided as a unit and encoded as separate groups (as separate hierarchies). The frequency domain difference signal collected into one or a plurality of parts is encoded in the difference information encoding unit 6 together with auxiliary information such as the amplitude of the difference value and the segment value for each frequency band, and the frequency domain difference signal encoded signal Then, it is supplied to the multiplexing / encoding information forming unit 7. The difference information encoding unit 6 will be described in more detail later.
[0026]
The multiplexed / encoded information forming unit 7 uses an auxiliary information from the psychoacoustic model 2 and a quantized frequency sample from the quantizer 3 as an acoustic signal encoded signal that can be reconstructed as a single unit at least in the first layer. Form. It may be transmitted as one encoded signal as it is, or may be multiplexed and transmitted together with the frequency domain differential signal encoded signal. In the case of multiplexing and transmitting, for example, the encoded signal of each layer as shown in FIG. 2 is transmitted as a bit string synthesized in time series.
[0027]
In FIG. 2, the frequency samples included in the first layer and the frequency samples in the second layer added as frequency domain difference information are encoded and multiplexed with the time series matched, so that the decoder Then, by decoding both pieces of information, a high-quality decoded signal can be obtained immediately. Note that the information of the frequency domain differential signal encoded signal in the second and lower layers is information missing in the first layer encoding unit, and the difference in the frequency domain generated when the original signal and the first layer are encoded. The signal is information obtained by encoding the importance of the bit, if necessary, auxiliary information such as time information and layer number, and the frequency domain difference signal itself. In FIG. 2, “SYNC” is synchronization information.
[0028]
In addition to the first layer acoustic signal encoded signal that can be restored alone, when the frequency domain difference information is encoded and transmitted, the identification signal is used in order to achieve time-series matching with the first layer in frame units. To be transmitted. Since it is desirable that the identification signal is different for each frame, for example, as shown in FIG. 3, the identification of the frequency domain difference signal encoded signal of the second layer or lower is assigned to the frame number of the encoded audio signal of the first layer. It can be added as a signal.
[0029]
When the first frame of the first layer acoustic signal encoded signal and the first frame number of the frequency domain difference signal encoded signal of the second layer or lower match and the frequency domain difference signal encoded signal always exists Need not add an identification signal indicating time-series matching.
[0030]
Furthermore, an error detection code is often included in the acoustic signal encoded signal forming the first layer. This error detection code can be used as an identification signal indicating a time-series match because the probability that the same code appears is very low except when the acoustic signal is silent. When such a method is adopted, as shown in FIG. 4, an error detection code created in the first layer may be added to the encoded signal in the second layer or lower. FIG. 4 shows an example in which a general 16-bit CRC error detection code is used as an identification signal, and the identification signal in the figure is represented by a hexadecimal value.
[0031]
Moreover, a sample encoding part is mentioned as a part by which the change for every flame | frame is anticipated in the acoustic signal encoding signal of a 1st hierarchy. As shown in FIG. 5, a sample group (bit string) in a specific frequency band or a bit string of a sample encoding unit that can be specified in a frame is adopted as an identification signal for time-series matching, and the frequencies in the second and lower layers By adding to the region difference signal encoded signal, the frequency sample can be corrected.
[0032]
The specific portion uses, for example, several bits from the jump destination determined from the top of the frame, or a bit string number bit that occupies the position of a specific frequency band sample.
FIG. 5 shows an example in which 16 bits after the 100th bit from the front end of the frame located in the frequency sample portion after quantization of the first layer are used as an identification number indicating a time axis match.
[0033]
The identification signal indicating the time coincidence does not always need to be added in the frame because the frame is updated with the flow of time, and can be inserted at a certain interval. Further, when the quality of the sound signal encoded signal of the silent section or the first layer is good, it is possible to avoid transmitting the frequency domain differential signal encoded signal of the second layer or lower. In order to cope with such a situation, if an identification signal indicating time-series matching is added to frames before and after an interval at which transmission is not performed, consistency with the encoded audio signal of the first layer is maintained.
[0034]
Using the means as described above, the multiplexing / encoding information forming unit 7 in FIG. 1 forms and outputs one or more encoded signals.
[0035]
Next, the difference information encoding unit 6 included in the acoustic signal encoding apparatus shown in FIG. 1 will be described in detail. When restoring an acoustic signal with high accuracy, that is, to create a lossless encoded signal having a calculation accuracy comparable to the original sound, all the frequency domain difference signals obtained by the difference calculator 5 are encoded and transmitted. Good.
[0036]
However, in order to improve the quality of the acoustic signal after decoding hierarchically step by step, it is more appropriate to divide the signal into a plurality of layers according to the importance of the bits and transmit. In addition, not only CD quality but also encoding audio signals having quality such as DVD Audio whose quantization bit length exceeds 16 bits, and some DVD Video, etc. It can be adapted to various forms of storage media and transmission paths.
[0037]
The frequency domain difference signal calculated by the difference calculator 5 is basically a quantization noise, although musical components that are not encoded remain, and can be handled as random noise (white noise). it can. However, the noise power varies from frequency band to frequency band. This is due to the difference in quantization accuracy individually set for each frequency band performed in the first-layer acoustic signal encoding processing.
[0038]
FIG. 6 shows a frequency versus energy change (power) characteristic diagram of an example of the frequency domain difference signal. This characteristic diagram shows on the frequency axis the power of the difference between the signal after frequency conversion in the acoustic signal encoding process and the frequency signal after quantization and inverse quantization. As shown in FIG. 6, this differential signal, which is quantization noise, differs for each frequency band and is not uniform.
[0039]
In order to encode the frequency domain differential signal, it is possible to improve the encoding efficiency by encoding the frequency domain differential signal together with the frequency domain differential signal using the amplitude information different for each frequency band as auxiliary information. Further, as shown in FIG. 7, the bit length sufficient to divide the frequency domain difference signal into predetermined frequency bands and re-quantize the signal having the maximum amplitude value in each frequency band (band 1 to band n). Therefore, the in-band difference signal may be encoded. In the example of FIG. 7, all frequency domain difference signals are treated as positive numbers, but are not limited to positive numbers. A negative signal can be handled by a sign bit and an amplitude value expressed by an absolute value.
[0040]
As another method, there is a method in which an intra-frame requantization value is obtained on the basis of the maximum amplitude value of the frequency domain difference signal in the frame, and is grouped and encoded according to the importance of bits over all the difference signals. At this time, the bit strings to be grouped may be in units of one bit, or a plurality of bits may be grouped and encoded. In this case, since the upper bits have a lot of redundancy, it is effective to use entropy coding such as Huffman coding or arithmetic coding.
[0041]
The example of FIG. 8 shows a case where 1 bit is grouped and encoded in units. The negative difference signal is bit-inverted except for the sign bit, thereby increasing the probability that the bit becomes “0”, and the effect of entropy coding is expected. It is possible to easily have a hierarchical structure by encoding each bit string group with an entropy encoder.
[0042]
Here, when encoding the frequency domain difference signal, it is not limited to either one of the encoding methods, for example, a band-specific encoding method including re-quantization information in the former frequency band for the upper bit group is adopted. However, a method of grouping and encoding the lower bit group according to the importance of the bits in the entire frame may be adopted, and they can be used in combination.
[0043]
Next, the acoustic signal decoding apparatus of the present invention will be described with reference to FIG. FIG. 9 shows a block diagram of an embodiment of an acoustic signal decoding apparatus according to the present invention. In the figure, the DeMUX / encoded information classification unit 8 receives one or a plurality of encoded signals output from the audio signal encoding device of FIG. The signals are classified into signals and frequency domain difference information encoded signals.
[0044]
That is, the DeMUX / encoded information classification unit 8 receives the frequency domain difference signal encoded signal of the second and lower layers in a state multiplexed with the first layer of the acoustic signal encoded signal that can be restored independently. At this time, the encoded signal is immediately separated, the quantized acoustic signal encoded signal included in the first layer is supplied to the inverse quantizer 9, and the frequency domain difference information encoded signal is supplied to the difference information decoding unit 10. To supply.
[0045]
On the other hand, the DeMUX / encoded information classifying unit 8 searches for an identification signal indicating the above-described time-series match when the encoded signal is input separately for each layer, and searches for the identification signal indicating the above-described time-series match. Since the encoded signals in the second and lower layers must be aligned, the frame synchronization function is provided, and the encoded signals in the first layer classified in a time-sequential state match each acoustic signal encoding method. Then, the signal is supplied to the inverse quantizer 9 to be inversely quantized and converted into frequency samples having an original amplitude value.
[0046]
On the other hand, the frequency domain differential signal encoded signal in the second layer and below is supplied to the differential information decoding unit 10 and similarly restored as a frequency domain differential signal having an original amplitude value. When there are a plurality of frequency domain differential signal encoded signals in the second layer and below, the frequency domain differential signal can be acquired by repeating the processing a plurality of times or by taking a multistage configuration. Here, when entropy coding is performed by the differential information encoding unit 6 in the acoustic signal encoder, the differential information decoding unit 10 decodes the entropy encoded signal.
[0047]
A plurality of frequency domain frequency samples (frequency spectrum inverse quantized values) obtained by inverse quantizing the acoustic signal encoded signal by the inverse quantizer 9 are supplied to the frequency sample adding unit 11 where differential information decoding is performed. It is added to the frequency domain difference information decoded by the converting unit 10 to correct the frequency sample of the encoded audio signal of the first layer, which probably contains a quantization error, into a more accurate frequency sample.
[0048]
Finally, the frequency sample taken out from the frequency sample adder 11 is supplied to the frequency inverse transformer 12 where the frequency domain signal is converted into a time domain signal and output as an acoustic signal. The frequency inverse transformation method used here is the same as the frequency transformation method used in the first-layer acoustic coding method.
[0049]
As described above, in the present embodiment, the encoding generated by the conventional acoustic encoding method is provided by including the frequency domain differential signal encoder that generates the encoded signal that corrects the frequency band quantization error. Compared to signals, it is possible to provide playback sound with better quality, and to transmit information flexibly while maintaining a certain quality according to the capacity of the storage medium and the status of the transmission band I can do things.
[0050]
Note that the present invention is not limited to the above-described embodiment, and the acoustic signal encoding device usually includes an arrangement of bits when encoding a frequency domain difference signal having a plurality of bits, and an amplitude value. A bit string generation unit and bit string identification information for selecting whether to encode as a bit string using a single bit having a plurality of difference samples in a state where the bit importance is aligned. .
[0051]
【The invention's effect】
As described above, according to the acoustic signal encoding device of the present invention, the frequency domain differential signal encoded signal obtained by encoding the quantization error included in the conventional acoustic signal encoding device as the frequency domain differential signal. Since the encoded signal having the hierarchical structure is generated by combining with the encoded signal of the first layer as the encoded signal of the second layer or lower, it is possible to generate a wideband from a narrowband transmission path or a low-capacity storage medium It can handle large-capacity media and can realize flexible information transmission.
[0052]
Further, according to the acoustic signal decoding device of the present invention, the frequency domain differential signal encoded signal obtained by encoding the quantization error included in the conventional acoustic signal encoding device as the frequency domain differential signal is converted into the acoustic signal. Since the encoded signal having a hierarchical structure combined with the encoded signal is decoded, in the decoding process, the frequency spectrum in the encoded audio signal of the first layer is corrected to a more accurate signal, and the original signal It is possible to provide a playback sound that is close to or equivalent to.
[Brief description of the drawings]
FIG. 1 is a block diagram of an embodiment of an acoustic signal encoding apparatus according to the present invention.
FIG. 2 is a diagram illustrating an example of a bit string of an audio signal encoded signal generated according to the present invention.
FIG. 3 is a diagram illustrating a method of adding a time-series matching identification signal using a frame number of the first layer.
FIG. 4 is a diagram illustrating a method for adding a time-series matching identification signal using an error detection code of the first layer.
FIG. 5 is a diagram illustrating a method of adding a time-series matching identification signal using a specific sample region portion in the first layer.
FIG. 6 is a power vs. frequency characteristic diagram of an example of a frequency domain difference signal.
FIG. 7 is a diagram illustrating an example in which frequency domain difference signals are grouped and encoded for each band.
FIG. 8 is a diagram illustrating an example in which frequency domain difference signals are grouped and coded according to bit importance.
FIG. 9 is a block diagram of an embodiment of an acoustic signal decoding apparatus according to the present invention.
FIG. 10 is a diagram showing a minimum audible limit characteristic which is a kind of human auditory characteristics.
[Explanation of symbols]
1 Frequency converter
2 Auditory psychological model
3 Quantizer
4,9 Inverse quantizer
5 Difference calculator
6 Difference information encoding unit
7 Multiplexing / encoding information forming unit
8 DeMUX / encoded information classifier
10 Differential information decoder
11 Frequency sample adder
12 Frequency inverse converter

Claims

An acoustic signal encoded signal forming the first layer obtained by quantizing frequency samples for a plurality of frequency regions of the acoustic signal and a signal for correcting a quantization error included in a higher layer than itself. An audio signal encoding device that outputs an encoded signal that is formed by encoding a second layer or lower layer obtained by encoding,
The supplied acoustic signal is divided into the plurality of frequency regions, and according to the quantization accuracy determined based on the weighting for each frequency region using auditory characteristics, the frequency samples of the acoustic signal for each frequency region An encoder that generates an acoustic signal encoded signal that forms the first layer by compressing the information by performing quantization;
An inverse quantization value of the frequency sample for each frequency region obtained by inverse quantization of the acoustic signal encoded signal generated in the encoder, and a pre-quantization value supplied to the encoder A frequency domain differential signal encoder that encodes a difference value between the acoustic signal and the frequency sample for each frequency domain, and generates a frequency domain differential signal encoded signal that is an encoded signal of the second layer or lower; An acoustic signal encoding device comprising:

The acoustic signal encoded signal generated by the encoder and the frequency domain differential signal encoded signal generated by the frequency domain differential signal encoder are multiplexed and output in a time-aligned state. The frequency domain difference in which the time-series information is added to the frequency-domain difference signal encoded signal by adding time-series information for making the first function to match the time series of the acoustic signal encoded signal to the frequency-domain difference signal encoded signal The acoustic signal encoding device according to claim 1, further comprising a multiplexing device having at least one function of a signal encoded signal and a second function for outputting the acoustic signal encoded signal separately. .

The acoustic signal is divided into a plurality of frequency domains, and frequency samples for each frequency domain of the acoustic signal are quantized according to quantization accuracy determined based on weighting for each frequency domain using auditory characteristics. The first-level acoustic signal encoded signal generated by the above, the inverse quantized value of the frequency sample for each frequency region obtained by dequantizing the acoustic signal encoded signal, and before quantization A frequency domain difference signal encoded signal of the second and lower layers, which is generated by encoding a difference value with the frequency sample for each frequency domain of the acoustic signal, is supplied, and the acoustic signal encoded signal Separating means for outputting the frequency domain difference signal encoded signal separately in a time-series matched state,
Dequantizing the acoustic signal encoded signal output by the separating unit to obtain a frequency sample dequantized value; and
A frequency domain differential signal decoding unit for decoding the frequency domain differential signal encoded signal output by the separating unit to obtain a frequency domain differential signal;
Frequency sample addition means for adding the frequency sample dequantized value and the frequency domain difference signal;
An acoustic signal decoding apparatus comprising: frequency inverse conversion means for converting a frequency sample output from the frequency sample addition means into a time signal.