JP4508490B2 - Encoding device and decoding device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To encode an audio signal with a small information quantity, while securing the sound quality. SOLUTION: The encoder is provided with a quantized spectrum string generating part for generating a quantized spectrum string, quantized by prescribed quantization precision, based on the audio signal and a circulation-type code vector quantization part for outputting a spectrum string code, including circulating position identification information showing how many reference spectrum strings have to be circulated in order to obtain the circulation-type quantization spectrum string, which is most similar to the quantization spectrum string.

Description

ここで、ｎｃｏｄｅは、ＥＰ_iが最大値ｍａｘとなる場合の循環値ｉである。
【００６３】
最大値判定ステップ２０３の判定の結果、ステップ２０２の計算結果が最大と判定される場合（最大値判定ステップ２０３のＹ）、保存ステップ２０４に進む。保存ステップ２０４において、計算したＥＰ_iを最大値ｍａｘとして更新する。また循環数ｉを循環型コードベクトル量子化部１０４内の格納部１１５に保存する。ただし、循環数ｉを循環型コードベクトル量子化部１０４内の任意の格納部に格納されてもよい。次いで、終端判定ステップ２０５に進む。
【００６４】
最大値判定ステップ２０３における判定の結果、ステップ２０２の計算結果が最大でない場合（最大値判定ステップ２０３のＮ）、終端判定ステップ２０５に進む。
【００６５】
なお、ｉが０の場合、初期化ステップ２０１において、最大値ｍａｘは０に初期化されているので、最大値判定ステップ２０３において、ＥＰ₀は最大と判定され、保存ステップ２０４において、ＥＰ₀は循環型コードベクトル量子化部１０４の格納部１１５に保存される。
【００６６】
終端判定ステップ２０５において、基準スペクトル列を循環させた複数の循環型量子化スペクトル列のすべての計算が終了したかどうかを判定する。すなわち、終端判定ステップ２０５において、ｉが最大値かどうかを判定する。
【００６７】
ｉが最大値である場合（ステップ２０５のＹ）、この動作を終了する。
【００６８】
ｉが最大値でない場合（ステップ２０５のＮ）、更新ステップ２０６において、ｉを１つインクリメントし、ステップ２０２に戻る。
【００６９】
次いで、循環型コードベクトル量子化部１０４内は、インクリメントされたｉに対してステップ２０２〜２０６の動作を繰り返す。
【００７０】
終了判定ステップ２０５において、基準スペクトル列を循環させた複数の循環型量子化スペクトル列の全てを計算した場合、すなわち、ｉが最大値まで到達して終了と判定される場合（終了判定ステップ２０５のＹ）、循環型コードベクトル量子化部１０４は、保存ステップ２０４において保存された最大ＥＰ_i値の循環値ｉをスペクトル列コードとして符号列転送部１０５に出力する。ここで、循環値ｉは、量子化スペクトル列と最も類似する循環型量子化スペクトル列が基準スペクトル列からｉ個循環させることにより得られることを意味し、本明細書において循環位置識別情報とよぶ。上記説明では、量子化スペクトル列ｚの要素は１６であるので、循環値ｉが取り得る値も１６個である。従って、循環位置識別情報は、４ビットの符号で表現することができる。
【００７１】
図４は、符号化列転送部１０５から外部装置１０８に出力される符号化列４００の構造の一例を示す。符号化列４００は、一般に、低い周波数帯域から高い周波数帯域の順番に符号化されたスペクトル列コードを含む。ここで、周波数帯域ｎに対応する符号化列４００、すなわち、スペクトル列コード４０１について説明する。
【００７２】
スペクトル列コード４０１は、周波数帯域ｎに対応する循環位置識別情報４０２を含む。ここで、上述したように、量子化スペクトル列の要素が１６個である場合、循環位置識別情報は４ビットで表現することができる。
【００７３】
また、スペクトル列コード４０１は、周波数帯域ｎの符号として、スペクトル正規化部１０１から出力されるスケールファクタと、スペクトル増幅部１０２から出力される補正ゲインとをさらに含んでもよい。
【００７４】
このように構成された符号化装置を用いることで、少ない固定ビットで符号化を行なうことができる。また、コードブック１０４ａの各コード（循環型量子化スペクトル列）が循環構造で作成されていることから、ｃ₀、ｃ₁、・・・、ｃ₁₅の１６個の要素のみを符号化装置および復号化装置が有すればよく（例えば、循環型コードベクトル量子化部１０４の格納部１１５）、それにより、要素を記憶するための格納部の容量を小さくすることができる。
【００７５】
なお、上記説明では、量子化スペクトル列および循環型量子化スペクトル列の要素の数が１６個の場合を説明したが、本発明はこれに限定されない。量子化スペクトル列および循環型量子化スペクトル列の要素の数は任意でよい。
【００７６】
なお、上記の説明では、量子化スペクトル列ｚと最も類似の循環型量子化スペクトル列を判定するために内積を使用した。しかし、本発明はこれに限定されない。例えば、内積変形関数を使用してもよい。本明細書において、内積変形関数とは、内積の係数に重み付けがしていある関数を意味する。具体的には、図２に示されるＥＰ_i計算ステップ２０２において、量子化スペクトル列ｚと循環型量子化スペクトル列を表すコードＰ_iとのベクトル内積変形関数ＥＰ_iを計算する。
【００７７】
ＥＰ₀’＝３ｃ₀ｚ₀＋２ｃ₁ｚ₁＋・・・＋０．３ｃ₁₅ｚ₁₅
ＥＰ₁’＝３ｃ₁₅ｚ₀＋２ｃ₀ｚ₁＋・・・＋０．３ｃ₁₄ｚ₁₅
ＥＰ₂’＝３ｃ₁₄ｚ₀＋２ｃ₁₅ｚ₁＋・・・＋０．３ｃ₁₃ｚ₁₅
・・・・・
ＥＰ₁₅’＝３ｃ₁ｚ₀＋２ｃ₂ｚ₁＋・・・＋０．３ｃ₀ｚ₁₅ （４）
式（４）は、ＥＰ_n’（ｎ＝０、１、２、．．．、１５）は、各コードＰ_n（ｎ＝０、１、２、．．．、１５）と、入力される量子化スペクトル列ｚの要素ｚ_n（ｎ＝０、１、２、．．．、１５）との内積変形関数値である。
【００７８】
内積においては、すべての係数が１であるため、ある周波数帯域内の周波数スペクトルデータの重要度も均一として計算される。しかし、式（４）に示すように、内積変形関数を用いると、ある周波数帯域内の周波数スペクトルデータの重要度を考慮することができる。例えば、より低い周波数の影響をより強く考慮する場合、より低い周波数成分のデータに対応する量子化スペクトル列の係数を大きくすればよい。
【００７９】
（実施の形態２）
次に、本発明の実施の形態２における符号化装置について説明する。実施の形態２における符号化装置は、循環型コードベクトル量子化部１０４の最大値判定ステップ２０３、保存ステップ２０４の動作が異なる点を除いて、実施の形態１で説明した符号化装置と同じである。
【００８０】
本実施の形態において最大値判定ステップ２０３（図２参照）は、例えば、式（５）を用いる。
【００８１】
【数２】

ここで、ａｂｓ（ ）は、（ ）内の絶対値を出力する関数であり、変数ｆａｃｅは、コードの値が反転されているかどうか示す。本明細書において、変数ｆａｃｅは、位相識別情報とよぶ。上記式において、変数ｆａｃｅ＝１の場合、図５に示されるコードブック１０４ｂからの１つのコードが、量子化スペクトル列に最も類似していることを意味し、変数ｆａｃｅ＝０の場合、図３に示されるコードブック１０４ａからの１つのコードが、量子化スペクトル列に最も類似していることを意味する。
【００８２】
図２に示される保存ステップ２０４において、ｎｃｏｄｅとｆａｃｅの値が保存される。
【００８３】
上記に示すように実施の形態２では、最大値判定ステップ２０３において、内積値ではなく、ＥＰ_i計算ステップ２０２において計算した値を用いて内積値の絶対値が最大となる循環値ｉを求める。これは、図３に示すコードブック１０４ａのコードを計算しながら、同時に、図５に示すコードブック１０４ｂのコードを計算しているのと同義である。なぜなら、コードブック１０４ｂのコードの各要素は、図３に示すコードの各要素の値の符号を全て反転させたものに対応するからである。ただし、式（５）に示すように計算することで、コードブック１０４ａ、１０４ｂのコードを順次計算する場合と比べて、大幅に計算時間を短縮することができる。
【００８４】
本実施の形態における循環型コードベクトル量子化部１０４のコードブックは、コードブック１０４ａ、１０４ｂに示されるコードを有する。コードブック１０４ｂの各コードは、コードブック１０４ａの各コードに対して符号をすべて反転させたものである。
【００８５】
図６は、符号化列転送部１０５から外部装置１０８に出力される符号化列６００の構造の一例を示す。符号化列６００は、一般に、低い周波数帯域から高い周波数帯域の順番に符号化されたスペクトル列コードを含む。ここで、周波数帯域ｎに対応する符号化列６００、すなわちスペクトル列コード６０１について説明する。
【００８６】
スペクトル列コード６０１は、周波数帯域ｎに対応する循環位置識別情報６０２と位相識別情報６０３とを含む。ここで、上述したように、量子化スペクトル列の要素が１６個である場合、循環位置識別情報は４ビットで表現することができ、位相識別情報は１ビットで表現することができる。
【００８７】
また、スペクトル列コード６０１は、周波数帯域ｎの符号として、スペクトル正規化部１０１から出力されるスケールファクタと、スペクトル増幅部１０２から出力される補正ゲインとをさらに含んでもよい。
【００８８】
このように構成された符号化装置では、実施の形態１と比較して計算量が最大値判定ステップ２０３の部分のみが大きくなるだけであり、また、スペクトル列コード６０１の観点から位相識別情報６０３に割り当てるビット数が１だけ増加するのみである。要素の数が１６の場合、本実施の形態は５ビットという少ない符号量で符号化を行なうことができる。
【００８９】
このように、本実施の形態によれば、少ない固定ビットで符号化を行なうことができる。また、コードブック１０４ｂに含まれる複数の循環型量子化スペクトル列は、コードブック１０４ａの基準スペクトル列を循環させた循環型量子化スペクトル列の全ての要素の符号を反転して作成されているため、ｃ₀、ｃ₁、・・・、ｃ₁₅の１６個の要素のみを符号化装置および復号化装置が有すればよく（例えば、循環型コードベクトル量子化部１０４の格納部１１５）、それにより、要素を記憶するための格納部の容量を小さくすることができる。
【００９０】
（実施の形態３）
次に、本発明の実施の形態３における符号化装置について説明する。実施の形態３における符号化装置は、コードブック１０４ｃと、ＥＰ_i計算ステップ２０２における動作が異なる点を除いて、実施の形態２で説明した符号化装置と同じである。
【００９１】
本実施の形態における循環型コードベクトル量子化部１０４のコードブック１０４ｃを図７に示す。本実施の形態におけるコードブック１０４ｃは、要素ｃ₀、ｃ₁、・・・、ｃ_n-1、ｃ_nのうち、一定の間隔で０の要素が配置されていることを特徴とする。図７に示す例では、ｃ₁、ｃ₂、ｃ₃、ｃ₅、ｃ₆、ｃ₇、ｃ₉、ｃ₁₀、ｃ₁₁、ｃ₁₃、ｃ₁₄、ｃ₁₅の値が０である。これに伴い、ＥＰ_i計算ステップ２０２での計算は、式（６）に示すように単純化される。
【００９２】
ＥＰ₀＝ｚ・Ｐ₀＝ｃ₀ｚ₀＋ｃ₄ｚ₄＋ｃ₈ｚ₈＋ｃ₁₂ｚ₁₂
ＥＰ₁＝ｚ・Ｐ₁＝ｃ₀ｚ₁＋ｃ₄ｚ₅＋ｃ₈ｚ₉＋ｃ₁₂ｚ₁₃
ＥＰ₂＝ｚ・Ｐ₂＝ｃ₀ｚ₂＋ｃ₄ｚ₆＋ｃ₈ｚ₁₀＋ｃ₁₂ｚ₁₄
・・・・・
ＥＰ₁₅＝ｚ・Ｐ₁₅＝ｃ₀ｚ₁₅＋ｃ₄ｚ₃＋ｃ₈ｚ₇＋ｃ₁₂ｚ₁₁ （６）
したがって、実施の形態２と比べて、ＥＰ_i計算ステップ２０２の計算量を４分の１に減らすことができる。
【００９３】
このように構成された符号化装置を用いることにより、少ない固定ビットで符号化を行なうことができる。また、コードブック１０４ｃの各コードが循環構造で作成されており、かつ、４つの要素（ｃ₀、ｃ₄、ｃ₈、ｃ₁₂）のみが値を有するので、符号化装置および復号化装置は４つの要素の値のみを有すればよく（例えば、循環型コードベクトル量子化部１０４の格納部１１５）、要素を記憶するための格納部の容量を小さくすることができる。
【００９４】
また、本実施の形態においても、実施の形態２において説明したのと同様に、循環型コードベクトル量子化部１０４のコードブックは、コードブック１０４ｃに示されるコードだけでなく、図８に示すコードブック１０４ｄに示されるコードを有してもよい。
【００９５】
なお、上記説明では、０の値を有する要素が連続する場合を説明したが、本発明はこれに限定されない。循環型量子化スペクトル列が有する任意の要素のうちの少なくとも１つ０の値を有する要素があればよい。
【００９６】
また、上記説明では、循環型量子化スペクトル列の要素の数が１６の場合を説明したが、本発明はこれに限定されない。循環型量子化スペクトル列の要素の数は任意の数であればよい。
【００９７】
（実施の形態４）
次に本発明の実施の形態４における符号化装置について説明する。本実施の形態における符号化装置は、ハフマン符号化部と符号化切換部とを備える点を除いて、実施の形態１で説明した符号化装置１００と同じ構成を有する。
【００９８】
図９は、実施の形態４における符号化装置９００を示す構成図である。符号化装置９００は、量子化スペクトル列生成部１１０と、循環型コードベクトル量子化部１０４と、符号化列転送部１０５と、ハフマン符号化部１０６と、符号化切換部１０７とを備える。量子化スペクトル列生成部１１０は、スペクトル正規化部１０１と、スペクトル増幅部１０２と、スペクトル量子化部１０３とを含む。
【００９９】
なお、量子化スペクトル列生成部１１０（スペクトル正規化部１０１と、スペクトル増幅部１０２と、スペクトル量子化部１０３）と、循環型コードベクトル量子化部１０４と、符号化列転送部１０５と、外部装置１０８とは、図１に示す符号化装置１００と同一であるので、それらの説明を省略する。
【０１００】
符号化切換部１０７は、スペクトル量子化部１０３で得られた量子化スペクトル列ｚをハフマン符号化するか、循環型量子化スペクトル列に変換するかを所定の条件に基づいて切り換える。このとき、符号化切換部１０７は、符号化方式を符号化転送部１０５に伝える。
【０１０１】
符号化切換部１０７が量子化スペクトル列ｚをハフマン符号部１０６に入力するように切り換えた場合、ハフマン符号化部１０６は、量子化スペクトル列ｚを変換してハフマン符号列を生成する。ハフマン符号化部１０６は、量子化スペクトルｚ_iの複数個をまとめてハフマン符号化する。ハフマン符号化によって符号化されたハフマン符号列を復号化する場合、復号化装置は、量子化スペクトルｚ_iを完全に復元することができる（ロスレス復号化）。
【０１０２】
符号化切換部１０７が量子化スペクトル列ｚを循環型コードベクトル量子化部１０４に入力するように切り換えた場合、循環型コードベクトル量子化部１０４は、量子化スペクトル列ｚを、量子化スペクトル列ｚに最も類似する循環型量子化スペクトル列に変換する。なお、量子化スペクトル列ｚに最も類似する循環型量子化スペクトル列は、実施形態１〜３に説明したように生成される。
【０１０３】
このような構成によれば、オーディオ信号が複数の周波数帯域に分割され、周波数帯域ごとに周波数スペクトル列を符号化する場合、符号化切換部１０７は、所定の条件に基づいて、量子化スペクトル列を循環型コードベクトル量子化部１０４か、ハフマン符号化部１０６に出力するかを切り換える。上記所定の条件とは、例えば、その量子化スペクトル列が、オーディオ信号の複数の周波数帯域のどの周波数帯域から派生したかによる。符号化に必要な情報量の配分が小さい場合（すなわち、その周波数帯域が受聴者の聴感に影響の少ない場合）、符号化切換部１０７は、量子化スペクトル列ｚを循環型コードベクトル量子化部１０４に出力するように切り換える。符号化に必要な情報量の配分が大きい場合（すなわち、その周波数帯域が受聴者の聴感に影響する場合）、符号化切換部１０７は、量子化スペクトル列ｚをハフマン符号化部１０６に出力するように切り換える。
【０１０４】
したがって、循環型コードベクトル量子化部１０４において、量子化スペクトル列ｚと、量子化スペクトル列ｚに最も類似する循環型量子化スペクトル列との類似度がそれほどよくなく、復号化の際に多少のロスが生じても、対象の周波数帯域は受聴者の聴感に影響の少ない。また、音質をある程度保ちながら、少ない情報量で符号化を行なうことができる。
【０１０５】
（実施の形態５）
次に、本発明の実施の形態５における符号化装置について説明する。実施の形態５における符号化装置は、コードブック１０４ｅの内容と、計算ステップ２０２における動作が異なる点を除いて、実施の形態３で説明した符号化装置と同じである。
【０１０６】
実施の形態１〜４において、コードブック１０４ａ、１０４ｂ、１０４ｃ、１０４ｄに含まれる複数の循環型量子化スペクトル列は、１つの基準スペクトル列を循環させた循環型量子化スペクトル列であった。しかし、本発明はこれに限定されない。本実施の形態では、コードブック１０４ｅが、複数の基準スペクトル列のそれぞれを循環させた循環型量子化スペクトル列を含む場合を説明する。
【０１０７】
図１０は、４つの基準スペクトル列を有するコードブック１０４ｅを模式的に示す。コードブック１０４ｅは、４つの基準スペクトル列のそれぞれを循環させた複数の循環型量子化スペクトル列を含むが、ここでは、簡略化のため、基準スペクトル列のみを示す。
【０１０８】
ここで、量子化スペクトル列ｚに最も類似する循環型量子化スペクトル列は、第２の基準スペクトル列を３つ循環させ、すべての要素を反転させたスペクトルが量子化スペクトル列ｚに最も類似するスペクトルであると仮定する。図１０に示されるように、第２の基準スペクトル列は｛２、０、０、０、−２、０、０、０、−１、０、０、０、１、０、０、０｝であり、これを３つ循環させて、すべての要素を反転させた循環型量子化スペクトル列は、｛０、２、０、０、０、１、０、０、０、−１、０、０、０、−２、０、０｝である。したがって、量子化スペクトル列ｚに最も類似する循環型量子化スペクトル列スペクトルは、｛０、２、０、０、０、１、０、０、０、−１、０、０、０、−２、０、０｝である。
【０１０９】
図１１は、この場合のスペクトル列コードを示す。ここで、ｃｏｄｅｂｏｏｋｉｄは基準スペクトル列識別情報であり、ｃｏｄｅ ｉｎｄｅｘは循環位置識別情報であり、ｐｈａｓｅは位相識別情報である。基準スペクトル列識別情報は、スペクトル列コードによって表される循環型量子化スペクトル列がコードブック１０４ｅのどの基準スペクトルから派生するかを示す情報である。循環位置識別情報は、スペクトル列コードによって表される循環型量子化スペクトル列がその基準スペクトル列をいくら循環することによって得られるかを示す情報である。位相識別情報は、スペクトル列コードによって表される循環型量子化スペクトル列が基準スペクトル列の全ての要素の符号を反転したスペクトル列に対応するか否かを示す情報である。
【０１１０】
図１２は、この場合に、符号化列転送部１０５から外部装置１０８に出力される符号化列１２００の構造の一例を示す。符号化列１２００は、一般に、低い周波数帯域から高い周波数帯域の順番に符号化されたスペクトル列コードを含む。ここで、周波数帯域ｎに対応する符号化列１２００、すなわちスペクトル列コード１２０１について説明する。
【０１１１】
スペクトル列コード１２０１は、周波数帯域ｎに対応する、追加情報１２０２と、基準スペクトル列識別情報１２０３と、循環位置識別情報１２０４と、位相識別情報１２０５とを含む。ここで、上述したように、コードブック１０４ｅが４個の基準スペクトル列を有する場合、基準スペクトル列識別情報１２０３は２ビットで表現される。また、周波数スペクトル列の要素が１６個である場合、循環位置識別情報１２０４は４ビットで表現される。また、位相識別情報１２０５は１ビットで表現される。
【０１１２】
上述した例の場合、基準スペクトル列識別情報１２０３と、循環位置識別情報１２０４と、位相識別情報１２０５は、１０進表記でそれぞれ、１、３、１である。したがって、これらは、２進表記で、それぞれ、０１、００１１、１と表現される。
【０１１３】
また、スペクトル列コード１２０１は、追加情報１２０２として、周波数帯域ｎの符号として、スペクトル正規化部１０１から出力されるスケールファクタ、および／または、スペクトル増幅部１０２から出力される補正ゲインとをさらに含んでもよい。
【０１１４】
（実施の形態６）
次に、本発明の実施の形態６として、復号化装置について説明する。本実施の形態の復号化装置は、実施の形態１〜３、５の符号化装置で符号化された符号化列を受け取り、上記符号化列を復号化して、オーディオ信号を得る。
【０１１５】
図１３は、本実施の形態による復号化装置１３００の構成を示すブロック図である。復号化装置１３００は、循環型コードベクトル逆量子化部１３０１と、スペクトル逆増幅部１３０２と、スペクトル逆正規化部１３０３と、符号化列入力部１３０４とを備える。
【０１１６】
符号化列入力部１３０４は、コードブック１３０７と、格納部１３０８とを含む。コードブック１３０７は、符号化列入力部１３０４に入力される符号化列を生成する際に使用されたコードと同じコードを含む。したがって、コードブック１３０７は、基準スペクトル列と、基準スペクトル列を循環させた複数の循環型量子化スペクトル列を含む。格納部１３０８は、コードブック１３０７のコードの各要素を格納する。
【０１１７】
ある場合、符号化列入力部１３０４は、符号化装置１００によって出力された符号化列を受け取る。符号化列入力部１３０４は、受け取った符号化列から循環型量子化スペクトル列を符号化した循環位置識別情報を抽出する。また、符号化列が、スペクトル正規化部１０１で用いられたスケールファクタおよび／またはスペクトル増幅部１０２で用いられた補正ゲインとを含む場合、符号化列入力部１３０４は、スケールファクタおよび／または補正ゲインを抽出する。
【０１１８】
循環型コードベクトル逆量子化部１３０１は、符号化列入力部１３０４から受け取った循環位置識別情報に基づいて、コードブック１３０７の循環型量子化スペクトル列から、該当する循環型量子化スペクトル列を量子化スペクトル列として選択し、スペクトル列コードのデータの復元を行なう。
【０１１９】
ここで、コードブック１３０７は、図３に示されるコードブック１０４ａと同じであると仮定する。例えば、循環位置識別情報の値ｉが１の場合のコードブック１３０７の循環型量子化スペクトル列は（ｃ15、ｃ0、ｃ1、．．．、ｃ13、ｃ14）であるので、量子化スペクトル列（ｃ15、ｃ0、ｃ1、．．．、ｃ13、ｃ14）がスペクトル逆増幅部１３０２に出力する。
【０１２０】
スペクトル逆増幅部１３０２は、スペクトル逆増幅部１３０２から入力されたスペクトル列に、符号化列入力部１３０４から受け取った補正ゲインを用いて逆増幅し、逆増幅スペクトル列を生成する。具体的には、符号化列入力部１３０４から受け取った補正ゲインがαであれば、１／α倍に増幅する。
【０１２１】
スペクトル逆正規化部１３０３は、逆増幅スペクトル列の各要素に対して、符号化列入力部１３０４から受け取ったスケールファクタを乗算し、各スペクトルにおける実際のレベル値に変換する。
【０１２２】
このようにスペクトル逆正規化部１３０３によって得られた実際のレベル値を示す１つの周波数帯域のスペクトルデータは、低域から高域へと並べられ、オーディオ信号の周波数スペクトルデータとして使用される。次いで、周波数時間変換部（図示せず）を用いて、周波数スペクトルデータを時間軸上のデータ、すなわち、ＰＣＭデータに変換する。さらに、このＰＣＭデータをＤＡ変換すると、アナログのオーディオ信号が生成される。
【０１２３】
なお、上記説明では、スペクトル列コードは循環位置識別情報のみを含む場合を説明したが、本発明はこれに限定されない。以下に、循環型コードベクトル逆量子化部１３０１は、循環位置識別情報だけでなく、基準スペクトル列識別情報と、位相識別情報とを含むスペクトル列コードから、スペクトル列コードが表す循環型量子化スペクトル列（すなわち、量子化スペクトル列）を生成する場合を説明する。
【０１２４】
図１４は、復号化装置１３００に用いられる循環型コードベクトル逆量子化部１３０１の動作を示すフローチャートである。循環型コードベクトル逆量子化部１３０１は、スペクトル列コード１２０１（図１２参照）に基づいて量子化スペクトル列を生成する。
【０１２５】
循環型コードベクトル逆量子化部１３０１は、ステップ１４０１において、基準スペクトル列識別情報に基づいて、循環型コードベクトル逆量子化部１３０１のコードブック１３０７に含まれる基準スペクトル列を特定する。この場合、コードブック１３０７は、図１０に示されるコードブック１０４ｄと同じである。
図１２に示されるように基準スペクトル列識別情報の値は１０進法表記で１（２進法表記で０１）であり、これは、基準スペクトル列識別情報が第２の基準スペクトル列であることを示す。
【０１２６】
ステップ１４０２において、循環型コードベクトル逆量子化部１３０１は、循環位置識別情報に基づいて、量子化スペクトル列を得るための基準スペクトル列循環量を取得する。循環位置識別情報は、図１２に示されるように、１０進法表記で３（２進法表記で００１１）である。
【０１２７】
循環型コードベクトル逆量子化部１３０１は、ステップ１４０３において、位相識別情報から位相の反転情報を取得する。位相識別情報は、図１２に示されるように、１０進法表記で１（２進法表記で１）である。
【０１２８】
上記の結果から、復号化装置１３００の循環型コードベクトル逆量子化部１３０１が、スペクトル列コード１２０１に基づいて生成される量子化スペクトル列は、｛０、２、０、０、０、１、０、０、０、−１、０、０、０、−２、０、０｝となる。
【０１２９】
（実施の形態７）
次いで、本発明の実施の形態７による復号化装置を説明する。図１５は、本実施の形態による復号化装置１５００の構成を示すブロック図である。復号化装置１５００は、図９に示される符号化装置９００によって符号化された符号化列から、オーディオ信号を得る。
【０１３０】
復号化装置１５００は、循環型コードベクトル逆量子化部１３０１と、スペクトル逆増幅部１３０２と、スペクトル逆正規化部１３０３と、符号化列入力部１３０４と、復号化切換部１３０５と、ハフマン逆量子化１３０６とを備える。循環型コードベクトル逆量子化部１３０１は、コードブック１３０７と、格納部１３０８とを含む。
【０１３１】
ただし、循環型コードベクトル逆量子化部１３０１と、スペクトル逆増幅部１３０２と、スペクトル逆正規化部１３０３と、符号化列入力部１３０４は、図１３に示される復号化装置１３００のものと同一であるで、説明を省略する。
【０１３２】
符号化列入力部１３０４は符号化列を受け取ると、ハフマン符号列または循環位置識別情報と、符号化方式とを抽出する。さらに、符号化列入力部１３０４は、補正ゲインおよびスケールファクタを取り出す。符号化列入力部１３０４は、符号化方式に関する情報を復号化切換部１３０５に出力する。復号化切換部１３０５は、符号化方式に基づいて、循環型コードベクトル逆量子化部１３０１またはハフマン逆量子化部１３０６を切り換える。循環位置識別情報は循環型コードベクトル逆量子化部１３０１に出力され、ハフマン符号列はハフマン逆量子化部１３０６に出力される。
【０１３３】
ハフマン逆量子化部１３０６は、ハフマンコードブックを格納する格納部１３０９を有する。復号化切換部１３０５がハフマン逆量子化部１３０６に切り換えて、ハフマン符号列を出力する場合、ハフマン逆量子化部１３０６が復号化処理を開始する。ハフマン逆量子化部１３０６は、ハフマンコードブック名とハフマン符号列が入力されると、ハフマン符号列に対応するインデックス値を読み出し、量子化スペクトル列を復元する。この場合、ロスレス復号化が実現される。
【０１３４】
復号化切換部１３０５が循環型コードベクトル逆量子化部１３０１に切り換えて、位置循環識別情報を出力する場合、実施の形態６で説明したのと同様な復号化処理が行なわれ、量子化スペクトル列が復元される。
【０１３５】
ハフマン逆量子化部１３０６または循環型コードベクトル逆量子化部１３０１によって生成された量子化スペクトル列は、実施の形態６において説明したように周波数スペクトルデータに変換される。
【０１３６】
次いで、周波数時間変換部（図示せず）は、上記の周波数スペクトルデータを時間軸上のデータ、すなわち、ＰＣＭデータに変換する。さらに、このＰＣＭデータをＤＡ変換すると、アナログのオーディオ信号が生成される。
【０１３７】
なお、上記説明では、循環型コードベクトル逆量子化部１３０１が、循環位置識別情報のみから量子化スペクトル列を生成する場合を説明したが、本発明はこれに限定されない。実施の形態６で説明したように、循環型コードベクトル逆量子化部１３０１は、循環位置識別情報だけでなく、基準スペクトル列識別情報と、位相識別情報とを含むスペクトル列コードから、スペクトル列コードが表す循環型量子化スペクトル列（すなわち、量子化スペクトル列）を生成してもよい。
【０１３８】
【発明の効果】
本発明による符号化装置は、量子化スペクトル列と最も類似する循環型量子化スペクトル列は基準スペクトル列をいくつ循環させたかを示す循環位置識別情報を含むスペクトル列コードを出力することにより、符号化時の情報量を少なくし、かつ、よりよい音質を得ることができる。
【０１３９】
この符号化装置は、従来の符号化方法と比べて、計算量が少なく、かつ、格納部の容量をを小さくすることができ、その結果、低いビットレートで効率的な符号化列を生成することができる。
【０１４０】
本発明による復号化装置は、基準スペクトル列を有し、基準スペクトル列と、上記基準スペクトル列をいくつ循環させることにより量子化スペクトル列が得られるかを示す循環位置識別情報とから、量子化スペクトル列を生成することにより、復号化装置が受け取る情報量が少なくでき、かつ、よりよい音質を効率的に得ることができる。
【図面の簡単な説明】
【図１】本発明の実施の形態１〜３、５における符号化装置の構成図。
【図２】本発明の実施の形態１〜５における符号化装置に用いられる循環型コードベクトル量子化部の動作を示すフローチャート。
【図３】実施の形態１の符号化装置に用いられるコードブック中の循環型量子化スペクトル列。
【図４】本発明の実施の形態１による符号化列の構造。
【図５】本発明の実施の形態２による符号化装置に用いられるコードブック中の循環型量子化スペクトル列。
【図６】本発明の実施の形態２による符号化列の構造。
【図７】本発明の実施の形態３による符号化装置に用いられるコードブック中の循環型量子化スペクトル列。
【図８】本発明の実施の形態３による符号化装置に用いられるコードブック中の別の循環型量子化スペクトル列。
【図９】本発明の実施の形態４による符号化装置の構成図。
【図１０】本発明の実施の形態５による符号化装置に用いられるコードブック中の複数の基準スペクトル列。
【図１１】本発明の実施の形態５によるスペクトル列コード。
【図１２】本発明の実施の形態５による符号化列の構造。
【図１３】本発明の実施の形態６による復号化装置の構成図。
【図１４】本発明の実施の形態６における復号化装置に用いられる循環型コードベクトル逆量子化部の動作を示すフローチャート。
【図１５】本発明の実施の形態７による復号化装置の構成図。
【図１６】従来の符号化装置の構成図。
【符号の説明】
１００ 符号化装置
１０１ スペクトル正規化部
１０２ スペクトル増幅部
１０３ スペクトル量子化部
１０４ 循環型コードベクトル量子化部
１０４ａ、１０４ｂ、１０４ｃ、１０４ｄ、１０４ｅ コードブック
１０５ 符号化列転送部
１０８ 外部装置
１１０ 量子化スペクトル列生成部
１１５ 格納部[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an encoding device and a decoding device. More specifically, the present invention relates to an encoding device and a decoding device that reduce the amount of information during encoding while maintaining the sound quality of an audio signal.
[0002]
[Prior art]
To date, a great number of methods for encoding and decoding audio signals including audio signals or music signals have been developed. In particular, recently, among them, IS13818-7 internationally standardized by ISO / IEC has been recognized and evaluated as a high sound quality and high efficiency encoding method. This encoding method is called AAC. In recent years, AAC is also adopted in a standard called MPEG4, and a system called MPEG4-AAC having several extended functions is defined with respect to IS13818-7. An example of the encoding process is described in INFOMATATE PART.
[0003]
Here, an encoding apparatus using a conventional encoding method will be described.
[0004]
FIG. 16 is a configuration diagram showing a conventional encoding device 1600. The encoding apparatus 1600 includes a spectrum normalization unit 1601, a spectrum amplification unit 1602, a spectrum quantization unit 1603, a Huffman encoding unit 1604, and an encoded sequence transfer unit 1605.
[0005]
The audio discrete signal (PCM data) obtained by sampling the audio signal is converted from the data on the time axis of the audio signal by using a method such as orthogonal transformation in a time frequency conversion unit (not shown). It is converted into frequency spectrum data of the signal. The data on the time axis of the audio signal is discrete data of the audio signal with respect to time, and the frequency spectrum data of the audio signal is discrete data of the audio signal with respect to frequency. The frequency spectrum data of the audio signal is input to the spectrum normalization unit 1601.
[0006]
The spectrum normalization unit 1601 receives a frequency spectrum sequence that is frequency spectrum data of one frequency band obtained by dividing the audio signal into a plurality of frequency bands, and specifies an average value of the frequency spectrum sequence using a scale factor. Generate a normalized spectral sequence that is normalized to the values in the range and expressed in floating point.
The scale factor is, for example, a multiplication coefficient in the power of 2.
[0007]
The spectrum amplifying unit 1602 receives the normalized spectrum sequence, and generates an amplified spectrum sequence corrected using the correction gain so that each value of the normalized spectrum sequence becomes a value in a specific range.
[0008]
The spectrum quantization unit 1603 receives the amplified spectrum sequence and generates a quantized spectrum sequence obtained by quantizing the amplified spectrum sequence using a predetermined conversion formula. In the case of the AAC method, the spectrum quantization unit 1603 rounds the spectrum data expressed in floating point to an integer value by quantization.
[0009]
The Huffman encoding unit 1604 converts the quantized spectrum sequence into a Huffman code sequence.
[0010]
The encoded sequence transfer unit 1605 outputs the scale factor output from the spectrum normalization unit 1601, the correction gain output from the spectrum amplification unit 1602, and the Huffman code sequence output from the Huffman encoding unit 1604 to the external device 1608. Forward to. The external device 1608 is, for example, a recording medium or a decoding device.
[0011]
[Problems to be solved by the invention]
In recent years, in order to reduce the amount of information at the time of encoding, it is desired to increase the compression rate of the audio signal.
[0012]
In the encoding apparatus 1600, the information compression performance depends on the Huffman encoding unit 1604. When encoding an audio signal with a high compression rate, that is, with a small amount of information in the encoding device 1600, the correction gain of the spectrum amplifying unit 1602 is controlled to reduce the value of the quantized spectrum sequence. The amount of information encoded by the encoding unit 1604 is reduced.
[0013]
However, when such an operation is performed, a very large value of amplitude (quantization value) zero appears in the frequency spectrum obtained by decoding the Huffman code string, and there is a problem that sufficient sound quality cannot be ensured.
[0014]
The present invention has been made in view of such conventional problems, and an encoding device that encodes an audio signal into an encoded sequence with a small amount of information while ensuring sound quality, and a frequency from the encoded sequence An object of the present invention is to realize a decoding device that decodes a spectrum sequence.
[0015]
[Means for Solving the Problems]
  The encoding device of the present invention includes a quantized spectrum sequence generating unit that generates a quantized spectrum sequence quantized with a predetermined quantization accuracy based on an audio signal, and a cyclic quantum sequence most similar to the quantized spectrum sequence. A cyclic code vector quantization unit that outputs a spectral sequence code including cyclic position identification information indicating how many times the standardized spectral sequence is obtained by circulating the reference spectral sequenceThe cyclic code vector quantization unit includes a codebook having a plurality of cyclic quantization spectrum sequences obtained by circulating the reference spectrum sequence, and the quantization spectrum among the plurality of cyclic quantization spectrum sequences. A cyclic quantization spectrum sequence having the maximum inner product with the sequence is set as a cyclic quantization spectrum sequence most similar to the quantization spectrum sequence.
  The encoding device of the present invention includes a quantized spectrum sequence generating unit that generates a quantized spectrum sequence quantized with a predetermined quantization accuracy based on an audio signal, and a cyclic quantum sequence most similar to the quantized spectrum sequence. A cyclic code vector quantization unit that outputs a spectral sequence code including cyclic position identification information indicating how many times the standardized spectral sequence is obtained by circulating the reference spectral sequence, and the cyclic code vector quantization unit includes: A codebook having a plurality of cyclic quantization spectrum sequences obtained by circulating the reference spectrum sequence, and the inner product deformation function with the quantization spectrum sequence of the plurality of circulation quantization spectrum sequences is maximum. The cyclic quantization spectrum sequence is a cyclic quantization spectrum sequence most similar to the quantization spectrum sequence.
[0016]
A Huffman encoding unit that outputs a Huffman code sequence obtained by converting the quantized spectrum sequence, and receives the quantized spectrum sequence, and the quantized spectrum sequence is either the cyclic code vector quantization unit or the Huffman encoding unit. And an encoding switching unit that switches based on a predetermined condition.
[0020]
The plurality of cyclic quantization spectrum sequences are:
P₀= (C₀, C₁, C₂,. . . , C_n-1, C_n)
P₁= (C_n, C₀, C₁,. . . , C_n-2, C_n-1)
P₂= (C_n-1, C_n, C₀,. . . , C_n-3, C_n-2)
...
P_n= (C₁, C₂, C_Three,. . . , C_n, C₀)
Where the reference spectral sequence is P₀And each of the plurality of cyclic quantization spectrum sequences has c₀, C₁, C₂,. . . , C_n-1, C_nAnd the number of elements in each of the plurality of cyclic quantization spectrum sequences may be n + 1.
[0021]
Each element c of the plurality of cyclic quantization spectrum sequences₀, C₁, C₂,. . . , C_n-1, C_nSome of them may be zero.
[0022]
Each element c of the plurality of cyclic quantization spectrum sequences₀, C₁, C₂,. . . , C_n-1, C_nThe elements at regular intervals may be zero.
[0023]
The quantized spectrum sequence generation unit generates the quantized spectrum sequence based on a frequency spectrum sequence that is spectrum data of one frequency band of the audio signal divided into a plurality of frequency bands, and the predetermined condition May be based on which frequency band of the plurality of frequency bands is the spectrum data.
[0024]
When the distribution of the information amount of the frequency band of the frequency spectrum sequence is large, the encoding switching unit may output the quantized spectrum sequence to the Huffman encoding unit.
[0025]
When the distribution of the information amount of the frequency band of the frequency spectrum sequence is small, the coding switching unit may output the quantized spectrum sequence to the cyclic code vector quantization unit.
[0026]
The code book may further include a second plurality of cyclic quantization spectrum sequences obtained by inverting the signs of all elements of the plurality of circulation quantization spectrum sequences.
[0027]
The plurality of cyclic quantized spectral sequences may include a cyclic quantized spectral sequence obtained by circulating each of a plurality of reference spectral sequences having the same number of elements as the number of elements of the quantized spectral sequence.
[0028]
  The decoding device of the present inventionA decoding device for decoding the encoded sequence output from the encoding device of the present invention described above,A reference spectrum sequence, and the reference spectrum sequence;Included in the coded sequenceBased on the spectral column codeSaidA cyclic code vector inverse quantization unit that generates a quantized spectrum sequence, and a spectrum inverse amplifying unit that receives the quantized spectrum sequence and deamplifies the quantized spectrum sequence using a correction gain to generate an amplified spectrum sequence And a spectrum denormalization unit that receives the amplified spectrum sequence and converts the amplified spectrum sequence using a scale factor to generate a frequency spectrum sequence, and the spectrum sequence code circulates the reference spectrum sequence in a number of ways. It includes cyclic position identification information indicating whether the quantized spectrum sequence can be obtained.
[0029]
  The decoding device of the present inventionA decoding apparatus for decoding the encoded sequence output from the encoding apparatus of the present invention described above, comprising the reference spectrum sequence, and the reference spectrum sequence and a spectrum sequence code included in the encoded sequence A cyclic code vector inverse quantization unit for generating the quantized spectrum sequence based on the code sequence, and included in the encoded sequenceA Huffman inverse quantization unit that receives a Huffman code sequence, converts the Huffman code sequence, and generates the quantized spectrum sequence;A spectrum inverse amplification unit that receives the quantized spectrum sequence, reverse-amplifies the quantized spectrum sequence using a correction gain to generate an amplified spectrum sequence, receives the amplified spectrum sequence, and uses the scale factor to generate the amplified spectrum A spectral denormalization unit that converts the sequence to generate a frequency spectrum sequence;AboveThe spectral sequence code included in the encoded sequence isThe cyclic code vector inverse quantization unitOutput toAboveThe Huffman code sequence included in the encoded sequence isHuffman inverse quantization sectionTo output the spectrum sequence code and the Huffman code sequence.Decoding switching section for switchingHave.
[0030]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an encoding device and a decoding device according to embodiments of the present invention will be described with reference to the drawings.
[0031]
(Embodiment 1)
FIG. 1 is a block diagram showing an encoding apparatus 100 according to the present invention. The encoding apparatus 100 includes a quantized spectrum sequence generation unit 110 that generates a quantized spectrum sequence quantized with a predetermined quantization accuracy based on an audio signal, and a circulation that outputs a spectrum sequence code based on the quantized spectrum sequence. Type code vector quantization section 104, and coded sequence transfer section 105 that transfers the output from quantized spectrum sequence generation section 110 and / or cyclic code vector quantization section 104 to external device 108.
[0032]
The quantized spectrum sequence generation unit 110 includes a spectrum normalization unit 101, a spectrum amplification unit 102, and a spectrum quantization unit 103.
[0033]
The quantized spectrum sequence generation unit 110 generates a quantized spectrum sequence from the audio signal as shown below.
[0034]
The audio discrete signal (PCM data) obtained by sampling the audio signal is converted from the data on the time axis of the audio signal by using a method such as orthogonal transform in a time-frequency conversion unit (not shown). It is converted into frequency spectrum data of the signal. Data on the time axis of the audio signal is discrete data of the audio signal with respect to time, and frequency spectrum data of the audio signal is discrete data of the audio signal with respect to frequency. The frequency spectrum data of the audio signal is input to the spectrum normalization unit 101.
[0035]
The spectrum normalization unit 101 receives the frequency spectrum sequence x, normalizes the received frequency spectrum sequence x using a scale factor, the average value or the maximum value of the frequency spectrum sequence to a value within a specific range, A normalized spectrum sequence y expressed in floating point is generated. The frequency spectrum sequence x is spectrum data of one frequency band among the audio signals divided into a plurality of frequency bands, and includes a predetermined number of data. The scale factor is, for example, a multiplication coefficient in the power of 2.
[0036]
The spectrum amplifying unit 102 generates an amplified spectrum sequence y · α amplified using the correction gain α so that each value of the normalized spectrum sequence y falls within a specific range. The correction gain α is used to correct a value in a specific range for each predetermined frequency band.
[0037]
The spectrum quantization unit 103 generates a quantized spectrum sequence z obtained by quantizing the amplified spectrum sequence y · α with a predetermined quantization accuracy using a predetermined conversion formula. In the case of the AAC scheme, the spectrum quantization unit 103 rounds the amplified spectrum sequence y · α expressed in floating point to an integer value by quantization, and generates a quantized spectrum sequence z having integer value data. .
[0038]
As will be apparent to those skilled in the art, in the quantized spectrum sequence generation unit 110, the frequency spectrum sequence, the normalized spectrum sequence, the amplified spectrum sequence, and the number of elements of the quantized spectrum sequence are equal.
[0039]
In the present embodiment, cyclic code vector quantization section 104 includes code book 104a and storage section 115. The code book 104a has a plurality of cyclic quantized spectral sequences obtained by circulating one reference spectral sequence having the same number of elements as the elements of the quantized spectral sequence z. The storage unit 115 stores the elements of the cyclic quantization spectrum sequence of the code book 104a.
[0040]
The cyclic code vector quantization unit 104 compares each of the plurality of cyclic quantization spectrum sequences with the quantized spectrum sequence z, and most of the quantization spectrum sequence z of the plurality of cyclic quantization spectrum sequences. Include a cyclic position identification information that determines a similar cyclic quantized spectral sequence and indicates how many cyclic quantized spectral sequences that are most similar to the quantized spectral sequence z can be obtained by cycling the reference spectral sequence Outputs the spectrum sequence code. As described above, the cyclic code vector quantization unit 104 converts the quantized spectral sequence z (more precisely, the cyclic quantized spectral sequence most similar to the quantized spectral sequence z) into a spectral sequence code and encodes it. The data is output to the transfer unit 105.
[0041]
The encoded sequence transfer unit 105 receives the scale factor output from the spectrum normalization unit 101, the correction gain output from the spectrum amplification unit 102, and the spectrum sequence code output from the cyclic code vector quantization unit 104. Transfer to external device 108. The external device 108 is, for example, a recording medium or a decoding device.
[0042]
The encoded sequence transfer unit 105 performs encoding based on a signal from any one of the spectrum normalization unit 101, the spectrum amplification unit 102, the spectrum quantization unit 103, and the cyclic code vector quantization unit 104. A sequence is generated, and a spectrum sequence code corresponding to the frequency spectrum sequence x input to the encoding device 100 is output to the external device 108. The encoded sequence transfer unit 105 may output only the spectral sequence code from the cyclic code vector quantization unit 104 to the external device 108.
[0043]
Hereinafter, the operation of the encoding apparatus 100 will be described in more detail.
[0044]
Audio discrete data (PCM data) is converted into frequency spectrum data at certain time intervals in a time-frequency converter (not shown). This frequency spectrum data is divided into a plurality of preset frequency bands, and a frequency spectrum sequence x is generated. The normalization unit 101 receives this frequency spectrum sequence x.
[0045]
The spectrum normalization unit 101 calculates the energy of the frequency spectrum sequence x input for each frequency band, and normalizes the calculated average value of the energy to a value within a specific range. The spectrum normalization unit 101 outputs the generated normalized spectrum sequence y to the spectrum amplification unit 102 and outputs the scale factor to the encoding / transfer unit 105.
[0046]
The spectrum amplifying unit 102 amplifies the normalized spectrum sequence y using the correction gain α so that each value is within a predetermined range, and generates an amplified spectrum sequence y · α. The spectrum amplification unit 102 outputs the amplified spectrum sequence y · α to the spectrum quantization unit 103, and outputs the correction gain α to the encoding transfer unit 105.
[0047]
The spectrum quantization unit 103 performs quantization using a predetermined conversion formula for the amplified spectrum sequence y · α. The conversion formula is, for example, the following formula (1).
[0048]
z_i= (Int) (y_i・ Α) (1)
In formula (1), z_iIs the i-th element of the quantized spectral sequence z and y_iIs the i-th element of the normalized spectral sequence y. α is a correction gain, and the correction gain is set for each divided frequency band. (Int) is a function that quantizes the value of the expression into an integer.
[0049]
By the equation (1), the amplified spectrum sequence y · α is converted into an integer-valued quantized spectrum sequence z. The spectrum quantization unit 103 outputs the quantized spectrum sequence z to the cyclic code vector quantization unit 104.
[0050]
As described above, the cyclic code vector quantization unit 104 selects a cyclic quantized spectral sequence that is most similar to the quantized spectral sequence z among a plurality of cyclic quantized spectral sequences obtained by circulating the reference spectral sequence. judge. The determination of which of the plurality of cyclic quantization spectrum sequences is most similar to the quantization spectrum sequence z is based on the inner product of each of the plurality of circulation quantization spectrum sequences and the quantization spectrum sequence z. Can be sought. This operation will be described in detail with reference to FIG.
[0051]
FIG. 2 is a flowchart showing the operation of the cyclic code vector quantization unit 104.
[0052]
When the quantized spectrum sequence z is input to the cyclic code vector quantization unit 104, i is initialized to 0 and the maximum value max to 0 in an initialization step 201.
[0053]
Then EP_iIn a calculation step 202, an inner product of the quantized spectrum sequence z input from the spectrum quantization unit 103 and the cyclic quantized spectrum sequence included in the code book 104a is calculated. Here, the number of elements of the quantized spectrum sequence z is the same as the number of elements of the cyclic quantized spectrum sequence. For example, when the number of elements in the quantized spectrum sequence z is 16, the number of elements in the cyclic quantized spectrum sequence is also 16. Thus, EP indicating the similarity between the quantized spectrum sequence z and the cyclic quantized spectrum sequence_i(I is the number of elements) is calculated. Hereinafter, a specific example in which the number of elements of each of the quantized spectrum sequence and the reference spectrum sequence, that is, the cyclic quantized spectrum sequence is 16, will be described.
[0054]
FIG. 3 shows a cyclic quantized spectrum sequence of the code book 104a used in the encoding apparatus of the present embodiment. The code representing the reference spectrum sequence is P₀, Code a plurality of cyclic quantized spectral sequences obtained by circulating the reference spectral sequence as a code P_nRepresented by
[0055]
In the present specification, the plurality of cyclic quantization spectrum sequences obtained by circulating the reference spectrum sequence includes the reference spectrum sequence itself. The reference spectrum sequence is a cyclic quantization spectrum sequence obtained by circulating the reference spectrum sequence 0 times.
[0056]
Code P shown in FIG._n(Here, n = 0, 1, 2,..., 15) has 16 elements.
[0057]
As shown in FIG. 3, the 0th code P₀Is {c₀, C₁, C₂,. . . , C₁₅}, The first code P₁Is {c₁₅, C₀, C₁,. . . , C₁₄}. As mentioned above, the code P₀Is the reference spectrum sequence and the first code P₁Is code P₀Each element is shifted to the right by one, and the first code P₁In element position 0, the 0th code P₀Element c at element position 15₁₅It becomes the structure which circulated. Second and subsequent code P₂~₁₅Similarly, the 0th code P₀Each of the elements is configured to circulate by the corresponding number. Therefore, the code P₀If each element of is found, another code P₁~₁₅Is code P₀It is determined uniquely by cycling.
[0058]
Note that the plurality of cyclic quantization spectrum sequences shown in FIG. 3 circulate on the right side, but the present invention is not limited to this, and may circulate on the left side.
[0059]
EP_iIn the calculation step 202, the code P representing the quantized spectrum sequence z and the cyclic quantized spectrum sequence_iVector inner product EP with_iCalculate
[0060]
EP₀= Z · P₀= C₀z₀+ C₁z₁+ ... + c₁₅z₁₅
EP₁= Z · P₁= C₁₅z₀+ C₀z₁+ ... + c₁₄z₁₅
EP₂= Z · P₂= C₁₄z₀+ C₁₅z₁+ ... + c₁₃z₁₅
...
EP₁₅= Z · P₁₅= C₁z₀+ C₂z₁+ ... + c₀z₁₅(2)
Equation (2) is expressed as EP_n(N = 0, 1, 2,..., 15)_n(N = 0, 1, 2,..., 15) and an input element z of the quantized spectrum sequence z_nThe inner product value (n = 0, 1, 2,..., 15).
[0061]
In the maximum value determination step 203, the cyclic code vector quantization unit 104 determines whether the calculation result of step 202 is the maximum of the results calculated so far. The maximum value determination step 203 uses, for example, the following expression (3).
[0062]
[Expression 1]

Where ncode is EP_iIs the circulation value i when becomes the maximum value max.
[0063]
As a result of the determination in the maximum value determination step 203, when the calculation result in step 202 is determined to be the maximum (Y in the maximum value determination step 203), the process proceeds to the storage step 204. In the saving step 204, the calculated EP_iIs updated as the maximum value max. Further, the circulation number i is stored in the storage unit 115 in the cyclic code vector quantization unit 104. However, the circulation number i may be stored in an arbitrary storage unit in the cyclic code vector quantization unit 104. Next, the process proceeds to the termination determination step 205.
[0064]
As a result of the determination in the maximum value determination step 203, when the calculation result in step 202 is not maximum (N in the maximum value determination step 203), the process proceeds to the termination determination step 205.
[0065]
If i is 0, the maximum value max is initialized to 0 in the initialization step 201. Therefore, in the maximum value determination step 203, EP₀Is determined to be the maximum, and in the save step 204, the EP₀Is stored in the storage unit 115 of the cyclic code vector quantization unit 104.
[0066]
In the end determination step 205, it is determined whether or not all the calculations of the plurality of cyclic quantization spectrum sequences obtained by circulating the reference spectrum sequence are completed. That is, in the termination determination step 205, it is determined whether i is the maximum value.
[0067]
If i is the maximum value (Y in step 205), this operation is terminated.
[0068]
If i is not the maximum value (N in Step 205), i is incremented by 1 in the update step 206, and the process returns to Step 202.
[0069]
Next, in the cyclic code vector quantization unit 104, the operations in steps 202 to 206 are repeated for the incremented i.
[0070]
In the end determination step 205, when all of the plurality of cyclic quantized spectrum sequences obtained by circulating the reference spectrum sequence are calculated, that is, when i reaches the maximum value and is determined to end (in the end determination step 205) Y) The cyclic code vector quantization unit 104 determines that the maximum EP stored in the storing step 204 is_iThe cyclic value i of the value is output to the code string transfer unit 105 as a spectrum string code. Here, the cyclic value i means that a cyclic quantized spectral sequence that is most similar to the quantized spectral sequence is obtained by circulating i pieces of cyclic quantized spectral sequences from the reference spectral sequence, and is referred to as cyclic position identification information in this specification. . In the above description, the number of elements of the quantized spectrum sequence z is 16, so that the circulation value i can take 16 values. Therefore, the cyclic position identification information can be expressed by a 4-bit code.
[0071]
FIG. 4 shows an example of the structure of the encoded sequence 400 output from the encoded sequence transfer unit 105 to the external device 108. The encoded sequence 400 generally includes a spectral sequence code encoded in order from a low frequency band to a high frequency band. Here, the encoded sequence 400 corresponding to the frequency band n, that is, the spectrum sequence code 401 will be described.
[0072]
The spectrum sequence code 401 includes circulation position identification information 402 corresponding to the frequency band n. Here, as described above, when there are 16 elements in the quantized spectrum sequence, the cyclic position identification information can be expressed by 4 bits.
[0073]
The spectrum sequence code 401 may further include a scale factor output from the spectrum normalization unit 101 and a correction gain output from the spectrum amplification unit 102 as codes of the frequency band n.
[0074]
By using the encoding apparatus configured as described above, encoding can be performed with a small number of fixed bits. In addition, since each code (cyclic quantization spectrum sequence) of the code book 104a is created in a cyclic structure, c₀, C₁, ..., c₁₅The encoding device and the decoding device only need to have 16 elements (for example, the storage unit 115 of the cyclic code vector quantization unit 104), thereby reducing the capacity of the storage unit for storing the elements. Can be small.
[0075]
In the above description, the case where the number of elements of the quantization spectrum sequence and the cyclic quantization spectrum sequence is 16 has been described, but the present invention is not limited to this. The number of elements of the quantization spectrum sequence and the cyclic quantization spectrum sequence may be arbitrary.
[0076]
In the above description, the inner product is used to determine the cyclic quantization spectrum sequence most similar to the quantization spectrum sequence z. However, the present invention is not limited to this. For example, an inner product deformation function may be used. In the present specification, the inner product deformation function means a function in which a coefficient of the inner product is weighted. Specifically, the EP shown in FIG._iIn the calculation step 202, the code P representing the quantized spectrum sequence z and the cyclic quantized spectrum sequence_iVector inner product deformation function EP_iCalculate
[0077]
EP₀'= 3c₀z₀+ 2c₁z₁+ ... + 0.3c₁₅z₁₅
EP₁'= 3c₁₅z₀+ 2c₀z₁+ ... + 0.3c₁₄z₁₅
EP₂'= 3c₁₄z₀+ 2c₁₅z₁+ ... + 0.3c₁₃z₁₅
...
EP₁₅'= 3c₁z₀+ 2c₂z₁+ ... + 0.3c₀z₁₅(4)
Equation (4) is expressed as EP_n′ (N = 0, 1, 2,..., 15) represents each code P_n(N = 0, 1, 2,..., 15) and an input element z of the quantized spectrum sequence z_nThe inner product deformation function value (n = 0, 1, 2,..., 15).
[0078]
In the inner product, since all the coefficients are 1, the importance of the frequency spectrum data within a certain frequency band is also calculated as uniform. However, as shown in Equation (4), the importance of the frequency spectrum data within a certain frequency band can be taken into account when the inner product deformation function is used. For example, when considering the influence of the lower frequency more strongly, the coefficient of the quantized spectrum sequence corresponding to the data of the lower frequency component may be increased.
[0079]
(Embodiment 2)
Next, an encoding apparatus according to Embodiment 2 of the present invention will be described. The encoding apparatus in the second embodiment is the same as the encoding apparatus described in the first embodiment except that the operations of the maximum value determination step 203 and the storage step 204 of the cyclic code vector quantization unit 104 are different. is there.
[0080]
In the present embodiment, the maximum value determination step 203 (see FIG. 2) uses, for example, Equation (5).
[0081]
[Expression 2]

Here, abs () is a function that outputs the absolute value in (), and the variable face indicates whether the value of the code is inverted. In this specification, the variable face is referred to as phase identification information. In the above equation, when the variable face = 1, it means that one code from the code book 104b shown in FIG. 5 is most similar to the quantized spectrum sequence, and when the variable face = 0, FIG. This means that one code from the code book 104a shown in Fig. 1 is most similar to the quantized spectrum sequence.
[0082]
In the storing step 204 shown in FIG. 2, the values of ncode and face are stored.
[0083]
As described above, in the second embodiment, in the maximum value determination step 203, not the inner product value but the EP._iUsing the value calculated in the calculation step 202, a circulation value i that maximizes the absolute value of the inner product value is obtained. This is equivalent to calculating the code of the code book 104b shown in FIG. 5 while calculating the code of the code book 104a shown in FIG. This is because each element of the code of the code book 104b corresponds to a code obtained by inverting all the signs of the values of the elements of the code shown in FIG. However, the calculation time can be significantly reduced by calculating as shown in Expression (5), compared to the case of sequentially calculating the codes of the codebooks 104a and 104b.
[0084]
The code book of cyclic code vector quantization section 104 in the present embodiment has codes shown in code books 104a and 104b. Each code of the code book 104b is obtained by inverting all the signs with respect to each code of the code book 104a.
[0085]
FIG. 6 shows an example of the structure of the encoded sequence 600 output from the encoded sequence transfer unit 105 to the external device 108. The encoded sequence 600 generally includes a spectral sequence code encoded in order from a low frequency band to a high frequency band. Here, the coded sequence 600 corresponding to the frequency band n, that is, the spectrum sequence code 601 will be described.
[0086]
The spectrum sequence code 601 includes circulation position identification information 602 and phase identification information 603 corresponding to the frequency band n. Here, as described above, when the number of elements of the quantized spectrum sequence is 16, the cyclic position identification information can be expressed by 4 bits, and the phase identification information can be expressed by 1 bit.
[0087]
The spectrum sequence code 601 may further include a scale factor output from the spectrum normalization unit 101 and a correction gain output from the spectrum amplification unit 102 as a code of the frequency band n.
[0088]
In the coding apparatus configured as described above, only the portion of the maximum value determination step 203 in the calculation amount is larger than that in the first embodiment, and the phase identification information 603 from the viewpoint of the spectrum sequence code 601. The number of bits allocated to is only increased by 1. When the number of elements is 16, the present embodiment can perform encoding with a small code amount of 5 bits.
[0089]
Thus, according to the present embodiment, encoding can be performed with a small number of fixed bits. In addition, the plurality of cyclic quantization spectrum sequences included in the code book 104b are created by inverting the signs of all elements of the cyclic quantization spectrum sequence obtained by circulating the reference spectrum sequence of the code book 104a. , C₀, C₁, ..., c₁₅The encoding device and the decoding device only need to have 16 elements (for example, the storage unit 115 of the cyclic code vector quantization unit 104), thereby reducing the capacity of the storage unit for storing the elements. Can be small.
[0090]
(Embodiment 3)
Next, an encoding apparatus according to Embodiment 3 of the present invention will be described. The encoding apparatus according to Embodiment 3 includes a code book 104c, an EP_iThe encoding apparatus is the same as that described in the second embodiment except that the operation in the calculation step 202 is different.
[0091]
FIG. 7 shows a code book 104c of the cyclic code vector quantization unit 104 in the present embodiment. The code book 104c in the present embodiment includes an element c₀, C₁, ..., c_n-1, C_nAmong them, 0 elements are arranged at regular intervals. In the example shown in FIG.₁, C₂, C_Three, C_Five, C₆, C₇, C₉, C_Ten, C₁₁, C₁₃, C₁₄, C₁₅The value of is 0. Along with this, EP_iThe calculation in calculation step 202 is simplified as shown in equation (6).
[0092]
EP₀= Z · P₀= C₀z₀+ C_Fourz_Four+ C₈z₈+ C₁₂z₁₂
EP₁= Z · P₁= C₀z₁+ C_Fourz_Five+ C₈z₉+ C₁₂z₁₃
EP₂= Z · P₂= C₀z₂+ C_Fourz₆+ C₈z_Ten+ C₁₂z₁₄
...
EP₁₅= Z · P₁₅= C₀z₁₅+ C_Fourz_Three+ C₈z₇+ C₁₂z₁₁(6)
Therefore, compared to the second embodiment, EP_iThe calculation amount of the calculation step 202 can be reduced to a quarter.
[0093]
By using the encoding apparatus configured as described above, encoding can be performed with a small number of fixed bits. In addition, each code of the code book 104c is created in a circular structure and includes four elements (c₀, C_Four, C₈, C₁₂) Only have values, the encoding device and the decoding device need only have the values of four elements (for example, the storage unit 115 of the cyclic code vector quantization unit 104) for storing the elements. The capacity of the storage unit can be reduced.
[0094]
Also in the present embodiment, as described in the second embodiment, the codebook of the cyclic code vector quantization unit 104 is not only the code shown in the codebook 104c but also the code shown in FIG. The code shown in the book 104d may be included.
[0095]
In the above description, the case where elements having a value of 0 are continuous has been described, but the present invention is not limited to this. There may be an element having a value of 0 of at least one of arbitrary elements included in the cyclic quantization spectrum sequence.
[0096]
In the above description, the case where the number of elements in the cyclic quantization spectrum sequence is 16 has been described, but the present invention is not limited to this. The number of elements in the cyclic quantization spectrum sequence may be an arbitrary number.
[0097]
(Embodiment 4)
Next, an encoding apparatus according to Embodiment 4 of the present invention will be described. The encoding apparatus in the present embodiment has the same configuration as that of encoding apparatus 100 described in Embodiment 1, except that it includes a Huffman encoding unit and an encoding switching unit.
[0098]
FIG. 9 is a configuration diagram showing an encoding apparatus 900 according to the fourth embodiment. The encoding apparatus 900 includes a quantized spectrum sequence generation unit 110, a cyclic code vector quantization unit 104, an encoded sequence transfer unit 105, a Huffman encoding unit 106, and an encoding switching unit 107. The quantized spectrum sequence generation unit 110 includes a spectrum normalization unit 101, a spectrum amplification unit 102, and a spectrum quantization unit 103.
[0099]
Note that the quantized spectrum sequence generation unit 110 (spectrum normalization unit 101, spectrum amplification unit 102, spectrum quantization unit 103), cyclic code vector quantization unit 104, encoded sequence transfer unit 105, external Since the apparatus 108 is the same as the encoding apparatus 100 shown in FIG. 1, description thereof is omitted.
[0100]
The encoding switching unit 107 switches whether the quantized spectrum sequence z obtained by the spectrum quantizing unit 103 is Huffman encoded or converted into a cyclic quantized spectrum sequence based on a predetermined condition. At this time, the encoding switching unit 107 informs the encoding transfer unit 105 of the encoding method.
[0101]
When the coding switching unit 107 switches to input the quantized spectrum sequence z to the Huffman code unit 106, the Huffman coding unit 106 converts the quantized spectrum sequence z to generate a Huffman code sequence. The Huffman encoding unit 106 performs quantization spectrum z_iHuffman coding is performed on a plurality of these. When decoding a Huffman code sequence encoded by Huffman coding, the decoding apparatus performs quantization spectrum z_iCan be completely restored (lossless decoding).
[0102]
When the encoding switching unit 107 switches to input the quantized spectrum sequence z to the cyclic code vector quantization unit 104, the cyclic code vector quantization unit 104 converts the quantized spectrum sequence z into the quantized spectrum sequence z. Convert to a circular quantized spectral sequence most similar to z. Note that the cyclic quantization spectrum sequence most similar to the quantization spectrum sequence z is generated as described in the first to third embodiments.
[0103]
According to such a configuration, when the audio signal is divided into a plurality of frequency bands and the frequency spectrum sequence is encoded for each frequency band, the encoding switching unit 107 performs the quantization spectrum sequence based on a predetermined condition. To be output to the cyclic code vector quantization unit 104 or the Huffman coding unit 106. The predetermined condition depends on, for example, which frequency band of the plurality of frequency bands of the audio signal the quantized spectrum sequence is derived from. When the distribution of the amount of information necessary for encoding is small (that is, when the frequency band has little influence on the listener's audibility), the encoding switching unit 107 converts the quantized spectrum sequence z into a cyclic code vector quantization unit. Switch to output to 104. When the distribution of the amount of information necessary for encoding is large (that is, when the frequency band affects the listener's audibility), encoding switching section 107 outputs quantized spectrum sequence z to Huffman encoding section 106. Switch as follows.
[0104]
Therefore, in the cyclic code vector quantization unit 104, the degree of similarity between the quantized spectral sequence z and the cyclic quantized spectral sequence that is most similar to the quantized spectral sequence z is not very good, and there is a slight difference in decoding. Even if a loss occurs, the target frequency band has little influence on the listener's audibility. In addition, encoding can be performed with a small amount of information while maintaining sound quality to some extent.
[0105]
(Embodiment 5)
Next, an encoding apparatus according to Embodiment 5 of the present invention will be described. The encoding apparatus in the fifth embodiment is the same as the encoding apparatus described in the third embodiment, except that the contents of the code book 104e and the operation in the calculation step 202 are different.
[0106]
In Embodiments 1 to 4, the plurality of cyclic quantization spectrum sequences included in the codebooks 104a, 104b, 104c, and 104d are cyclic quantization spectrum sequences obtained by circulating one reference spectrum sequence. However, the present invention is not limited to this. In the present embodiment, a case will be described in which codebook 104e includes a cyclic quantization spectrum sequence obtained by circulating each of a plurality of reference spectrum sequences.
[0107]
FIG. 10 schematically shows a code book 104e having four reference spectrum sequences. The code book 104e includes a plurality of cyclic quantization spectrum sequences obtained by circulating each of the four reference spectrum sequences, but only the reference spectrum sequence is shown here for the sake of simplicity.
[0108]
Here, the cyclic quantized spectral sequence most similar to the quantized spectral sequence z has three second reference spectral sequences circulated, and the spectrum with all elements inverted is most similar to the quantized spectral sequence z. Assume that it is a spectrum. As shown in FIG. 10, the second reference spectrum sequence is {2, 0, 0, 0, -2, 0, 0, 0, -1, 0, 0, 0, 1, 0, 0, 0}. The cyclic quantized spectral sequence obtained by circulating three of them and inverting all the elements is {0, 2, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0, -2, 0, 0}. Therefore, the circular quantization spectrum sequence most similar to the quantization spectrum sequence z is {0, 2, 0, 0, 0, 1, 0, 0, 0, −1, 0, 0, 0, −2 , 0, 0}.
[0109]
FIG. 11 shows a spectrum sequence code in this case. Here, codebook is reference spectrum sequence identification information, code index is circulation position identification information, and phase is phase identification information. The reference spectrum sequence identification information is information indicating which reference spectrum of the code book 104e the cyclic quantized spectrum sequence represented by the spectrum sequence code is derived from. The cyclic position identification information is information indicating how much the cyclic quantized spectral sequence represented by the spectral sequence code is obtained by circulating the reference spectral sequence. The phase identification information is information indicating whether or not the cyclic quantization spectrum sequence represented by the spectrum sequence code corresponds to a spectrum sequence obtained by inverting the signs of all the elements of the reference spectrum sequence.
[0110]
FIG. 12 shows an example of the structure of the encoded sequence 1200 output from the encoded sequence transfer unit 105 to the external device 108 in this case. The encoded sequence 1200 generally includes a spectral sequence code encoded in order from a low frequency band to a high frequency band. Here, the encoded sequence 1200 corresponding to the frequency band n, that is, the spectrum sequence code 1201 will be described.
[0111]
The spectrum sequence code 1201 includes additional information 1202, reference spectrum sequence identification information 1203, circulation position identification information 1204, and phase identification information 1205 corresponding to the frequency band n. Here, as described above, when the code book 104e has four reference spectrum sequences, the reference spectrum sequence identification information 1203 is expressed by 2 bits. In addition, when there are 16 elements in the frequency spectrum sequence, the circulating position identification information 1204 is expressed by 4 bits. Further, the phase identification information 1205 is expressed by 1 bit.
[0112]
In the case of the above-described example, the reference spectrum sequence identification information 1203, the circulation position identification information 1204, and the phase identification information 1205 are 1, 3, and 1, respectively, in decimal notation. Therefore, these are expressed as 01, 0011, 1 in binary notation, respectively.
[0113]
The spectrum sequence code 1201 further includes, as additional information 1202, a scale factor output from the spectrum normalization unit 101 and / or a correction gain output from the spectrum amplification unit 102 as a code of the frequency band n. But you can.
[0114]
(Embodiment 6)
Next, a decoding device will be described as a sixth embodiment of the present invention. The decoding apparatus according to the present embodiment receives the encoded sequence encoded by the encoding apparatuses according to the first to third and fifth embodiments, decodes the encoded sequence, and obtains an audio signal.
[0115]
FIG. 13 is a block diagram showing a configuration of decoding apparatus 1300 according to the present embodiment. Decoding apparatus 1300 includes cyclic code vector inverse quantization section 1301, spectrum inverse amplification section 1302, spectrum inverse normalization section 1303, and encoded sequence input section 1304.
[0116]
The encoded string input unit 1304 includes a code book 1307 and a storage unit 1308. The code book 1307 includes the same code as the code used when generating the encoded sequence input to the encoded sequence input unit 1304. Therefore, the code book 1307 includes a reference spectrum sequence and a plurality of cyclic quantization spectrum sequences obtained by circulating the reference spectrum sequence. The storage unit 1308 stores each element of the code of the code book 1307.
[0117]
In some cases, the encoded sequence input unit 1304 receives the encoded sequence output by the encoding apparatus 100. The encoded sequence input unit 1304 extracts the cyclic position identification information obtained by encoding the cyclic quantized spectral sequence from the received encoded sequence. Further, when the encoded sequence includes the scale factor used in the spectrum normalization unit 101 and / or the correction gain used in the spectrum amplification unit 102, the encoded sequence input unit 1304 displays the scale factor and / or correction. Extract the gain.
[0118]
Based on the cyclic position identification information received from the coded sequence input unit 1304, the cyclic code vector inverse quantization unit 1301 quantizes the corresponding cyclic quantized spectral sequence from the cyclic quantized spectral sequence of the codebook 1307. The selected spectral sequence is selected and the spectral sequence code data is restored.
[0119]
  Here, the code book 1307 is the code shown in FIG.bookAssume that it is the same as 104a. For example, when the cyclic position identification information value i is 1, the cyclic quantized spectral sequence of the codebook 1307 is (c15, c0, c1,..., C13, c14), so that the quantized spectral sequence (c15 , C 0, c 1,..., C 13, c 14) are output to the spectrum inverse amplification unit 1302.
[0120]
The spectrum inverse amplification unit 1302 inversely amplifies the spectrum sequence input from the spectrum inverse amplification unit 1302 using the correction gain received from the encoded sequence input unit 1304 to generate an inverse amplification spectrum sequence. Specifically, if the correction gain received from the coded sequence input unit 1304 is α, it is amplified by 1 / α.
[0121]
The spectrum inverse normalization unit 1303 multiplies each element of the deamplified spectrum sequence by the scale factor received from the encoded sequence input unit 1304, and converts it into an actual level value in each spectrum.
[0122]
Thus, the spectrum data of one frequency band indicating the actual level value obtained by the spectrum inverse normalization unit 1303 is arranged from the low band to the high band, and is used as the frequency spectrum data of the audio signal. Next, a frequency time conversion unit (not shown) is used to convert the frequency spectrum data into data on the time axis, that is, PCM data. Further, when this PCM data is D / A converted, an analog audio signal is generated.
[0123]
In the above description, the case where the spectrum sequence code includes only the cyclic position identification information has been described, but the present invention is not limited to this. Hereinafter, the cyclic code vector inverse quantization unit 1301 uses the cyclic quantization spectrum represented by the spectrum sequence code from the spectrum sequence code including not only the cyclic position identification information but also the reference spectrum sequence identification information and the phase identification information. A case where a sequence (that is, a quantized spectrum sequence) is generated will be described.
[0124]
FIG. 14 is a flowchart showing the operation of the cyclic code vector inverse quantization unit 1301 used in the decoding apparatus 1300. The cyclic code vector inverse quantization unit 1301 generates a quantized spectrum sequence based on the spectrum sequence code 1201 (see FIG. 12).
[0125]
In step 1401, the cyclic code vector inverse quantization unit 1301 specifies a reference spectrum sequence included in the code book 1307 of the cyclic code vector inverse quantization unit 1301 based on the reference spectrum sequence identification information. In this case, the code book 1307 is the same as the code book 104d shown in FIG.
As shown in FIG. 12, the value of the reference spectrum sequence identification information is 1 in decimal notation (01 in binary notation). This is because the reference spectrum sequence identification information is the second reference spectrum sequence. Indicates.
[0126]
In step 1402, the cyclic code vector inverse quantization unit 1301 acquires a reference spectral sequence circulation amount for obtaining a quantized spectral sequence based on the cyclic position identification information. As shown in FIG. 12, the circulating position identification information is 3 in decimal notation (0011 in binary notation).
[0127]
In step 1403, cyclic code vector inverse quantization section 1301 obtains phase inversion information from phase identification information. As shown in FIG. 12, the phase identification information is 1 in decimal notation (1 in binary notation).
[0128]
From the above result, the quantized spectrum sequence generated by the cyclic code vector inverse quantization unit 1301 of the decoding apparatus 1300 based on the spectrum sequence code 1201 is {0, 2, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0, -2, 0, 0}.
[0129]
(Embodiment 7)
Next, a decoding apparatus according to Embodiment 7 of the present invention will be described. FIG. 15 is a block diagram showing a configuration of decoding apparatus 1500 according to the present embodiment. The decoding apparatus 1500 obtains an audio signal from the encoded sequence encoded by the encoding apparatus 900 shown in FIG.
[0130]
Decoding apparatus 1500 includes cyclic code vector inverse quantization section 1301, spectrum inverse amplification section 1302, spectrum inverse normalization section 1303, coded sequence input section 1304, decoding switching section 1305, Huffman inverse quantum. 1306. The cyclic code vector inverse quantization unit 1301 includes a code book 1307 and a storage unit 1308.
[0131]
However, cyclic code vector inverse quantization section 1301, spectrum inverse amplification section 1302, spectrum inverse normalization section 1303, and encoded sequence input section 1304 are the same as those of decoding apparatus 1300 shown in FIG. Therefore, the description is omitted.
[0132]
When the encoded sequence input unit 1304 receives the encoded sequence, the encoded sequence input unit 1304 extracts a Huffman code sequence or cyclic position identification information and an encoding method. Further, the encoded sequence input unit 1304 extracts the correction gain and the scale factor. The coded sequence input unit 1304 outputs information related to the coding scheme to the decoding switching unit 1305. Decoding switching section 1305 switches cyclic code vector inverse quantization section 1301 or Huffman inverse quantization section 1306 based on the coding scheme. The cyclic position identification information is output to the cyclic code vector inverse quantization unit 1301, and the Huffman code string is output to the Huffman inverse quantization unit 1306.
[0133]
The Huffman inverse quantization unit 1306 includes a storage unit 1309 that stores a Huffman codebook. When the decoding switching unit 1305 switches to the Huffman inverse quantization unit 1306 and outputs a Huffman code string, the Huffman inverse quantization unit 1306 starts decoding processing. When the Huffman codebook name and the Huffman code string are input, the Huffman inverse quantization unit 1306 reads an index value corresponding to the Huffman code string and restores the quantized spectrum string. In this case, lossless decoding is realized.
[0134]
When decoding switching section 1305 switches to cyclic code vector inverse quantization section 1301 and outputs position circulation identification information, decoding processing similar to that described in Embodiment 6 is performed, and a quantized spectrum sequence is performed. Is restored.
[0135]
The quantized spectrum sequence generated by the Huffman inverse quantization unit 1306 or the cyclic code vector inverse quantization unit 1301 is converted into frequency spectrum data as described in the sixth embodiment.
[0136]
Next, a frequency time conversion unit (not shown) converts the frequency spectrum data into data on the time axis, that is, PCM data. Further, when this PCM data is D / A converted, an analog audio signal is generated.
[0137]
In the above description, the case where the cyclic code vector inverse quantization unit 1301 generates the quantized spectrum sequence only from the cyclic position identification information has been described, but the present invention is not limited to this. As described in the sixth embodiment, the cyclic code vector inverse quantization unit 1301 obtains the spectrum sequence code from the spectrum sequence code including not only the cyclic position identification information but also the reference spectrum sequence identification information and the phase identification information. A cyclic quantization spectrum sequence represented by (ie, a quantization spectrum sequence) may be generated.
[0138]
【The invention's effect】
The encoding apparatus according to the present invention outputs a spectral sequence code including cyclic position identification information indicating how many reference spectral sequences have been circulated from a cyclic quantized spectral sequence that is most similar to the quantized spectral sequence. The amount of information at the time can be reduced, and better sound quality can be obtained.
[0139]
This encoding apparatus has a smaller amount of calculation and a smaller storage capacity than the conventional encoding method, and as a result, generates an efficient encoded sequence at a low bit rate. be able to.
[0140]
The decoding apparatus according to the present invention has a reference spectrum sequence, and a quantized spectrum from the reference spectrum sequence and cyclic position identification information indicating how many times the reference spectrum sequence is circulated to obtain a quantized spectrum sequence. By generating a sequence, the amount of information received by the decoding device can be reduced, and better sound quality can be obtained efficiently.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of an encoding device according to Embodiments 1 to 3 of the present invention.
FIG. 2 is a flowchart showing an operation of a cyclic code vector quantization unit used in the encoding apparatus according to Embodiments 1 to 5 of the present invention.
FIG. 3 shows a cyclic quantized spectrum sequence in a code book used in the coding apparatus according to the first embodiment.
FIG. 4 shows a structure of a coded sequence according to Embodiment 1 of the present invention.
FIG. 5 shows a cyclic quantized spectrum sequence in a codebook used in an encoding apparatus according to Embodiment 2 of the present invention.
FIG. 6 shows a structure of a coded sequence according to Embodiment 2 of the present invention.
FIG. 7 shows a cyclic quantized spectrum sequence in a codebook used in an encoding apparatus according to Embodiment 3 of the present invention.
FIG. 8 shows another cyclic quantized spectrum sequence in the codebook used in the encoding apparatus according to Embodiment 3 of the present invention.
FIG. 9 is a configuration diagram of an encoding apparatus according to Embodiment 4 of the present invention.
FIG. 10 shows a plurality of reference spectrum sequences in a code book used in an encoding apparatus according to Embodiment 5 of the present invention.
FIG. 11 shows a spectrum sequence code according to the fifth embodiment of the present invention.
FIG. 12 shows a structure of a coded sequence according to the fifth embodiment of the present invention.
FIG. 13 is a configuration diagram of a decoding apparatus according to Embodiment 6 of the present invention.
FIG. 14 is a flowchart showing an operation of a cyclic code vector inverse quantization unit used in the decoding apparatus according to Embodiment 6 of the present invention.
FIG. 15 is a configuration diagram of a decoding apparatus according to a seventh embodiment of the present invention.
FIG. 16 is a configuration diagram of a conventional encoding device.
[Explanation of symbols]
100 Encoder
101 Spectral normalization part
102 Spectral amplifier
103 Spectral quantization unit
104 Cyclic code vector quantization section
104a, 104b, 104c, 104d, 104e Codebook
105 Encoded sequence transfer unit
108 External device
110 Quantized spectrum sequence generator
115 Storage unit

Claims (13)

A quantized spectrum sequence generating unit that generates a quantized spectrum sequence quantized with a predetermined quantization accuracy based on the audio signal;
A cyclic code vector quantization unit that outputs a spectral sequence code including cyclic position identification information indicating how many cyclic spectral sequences that are most similar to the quantized spectral sequence are obtained by circulating a reference spectral sequence; equipped with a,
The cyclic code vector quantization unit includes:
A codebook having a plurality of cyclic quantized spectral sequences obtained by circulating the reference spectral sequence;
Among the plurality of cyclic quantization spectral sequences, a cyclic quantization spectral sequence having a maximum inner product with the quantized spectral sequence is set as a cyclic quantized spectral sequence most similar to the quantized spectral sequence,
It marks Goka apparatus. A quantized spectrum sequence generating unit that generates a quantized spectrum sequence quantized with a predetermined quantization accuracy based on the audio signal;
A cyclic code vector quantization unit that outputs a spectral sequence code including cyclic position identification information indicating how many cyclic spectral sequences that are most similar to the quantized spectral sequence are obtained by circulating a reference spectral sequence; With
The cyclic code vector quantization unit includes:
A codebook having a plurality of cyclic quantized spectral sequences obtained by circulating the reference spectral sequence;
Among the plurality of cyclic quantized spectral sequences, a cyclic quantized spectral sequence showing the maximum inner product deformation function with the quantized spectral sequence is a cyclic quantized spectral sequence most similar to the quantized spectral sequence, To
It marks Goka apparatus. A Huffman encoder that outputs a Huffman code sequence obtained by converting the quantized spectrum sequence;
An encoding switching unit that receives the quantized spectrum sequence and switches whether to output the quantized spectrum sequence to the cyclic code vector quantization unit or the Huffman encoding unit based on a predetermined condition; The encoding apparatus according to claim 1 or 2 . The plurality of cyclic quantization spectrum sequences are:
P 0 = (c ₀ , c ₁ , c ₂ ,..., C _n−1 , c _n )
P1 = (c _n , c ₀ , c ₁ ,..., C _n−2 , c _n−1 )
P2 = (c _n-1 , c _n , c ₀ , ..., c _n-3 , c _n-2 )
...
Pn = (c ₁ , c ₂ , c ₃ ,..., C _n , c ₀ )
Expressed in
Here, the reference spectrum sequence is P 0, and the elements of each of the plurality of cyclic quantization spectrum sequences are c ₀ , c ₁ , c ₂ ,. . . , C _n-1 , c _n ,
The encoding apparatus according to claim 1 or 2 , wherein the number of elements of each of the plurality of cyclic quantization spectrum sequences is n + 1. Each element c ₀ , c ₁ , c ₂ ,. . . , Some of the c _n-1, c n is 0, the coding apparatus according to claim 4. Each element c ₀ , c ₁ , c ₂ ,. . . , The elements of constant spacing of the c _n-1, c n is 0, the coding apparatus according to claim 4. The quantized spectrum sequence generating unit generates the quantized spectrum sequence based on a frequency spectrum sequence that is spectrum data of one frequency band of the audio signal divided into a plurality of frequency bands,
The encoding apparatus according to claim 3 , wherein the predetermined condition is based on which frequency band of the plurality of frequency bands the frequency spectrum sequence is. The encoding device according to claim 7 , wherein when the information amount distribution of the frequency band of the frequency spectrum sequence is large, the encoding switching unit outputs the quantized spectrum sequence to the Huffman encoding unit. The encoding apparatus according to claim 7 , wherein when the distribution of the information amount of the frequency band of the frequency spectrum sequence is small, the encoding switching unit outputs the quantization spectrum sequence to the cyclic code vector quantization unit. The codebook further comprising a plurality of the second plurality of circulating quantized spectral sequence obtained by reversing the sign of the positive and negative all elements included in the circulating quantized spectral sequence, to claim 1 or claim 2 The encoding device described. The plurality of cyclic quantized spectral sequences includes a cyclic quantized spectral sequence obtained by circulating each of a plurality of reference spectral sequences having the same number of elements as the number of elements of the quantized spectral sequence. The encoding device according to claim 1 or 2 . A decoding device that decodes an encoded sequence output from the encoding device according to claim 1 or 2,
Has the reference spectral sequence, and the reference spectral sequence, and the based on the spectral sequence code contained in the encoded data stream, circulating code vector inverse quantization unit which generates the quantized spectral sequence,
A spectrum inverse amplification unit that receives the quantized spectrum sequence and deamplifies the quantized spectrum sequence using a correction gain to generate an amplified spectrum sequence;
A spectrum denormalization unit that receives the amplified spectrum sequence and converts the amplified spectrum sequence using a scale factor to generate a frequency spectrum sequence;
The decoding apparatus, wherein the spectrum sequence code includes cyclic position identification information indicating how many times the reference spectrum sequence is circulated to obtain the quantized spectrum sequence. A decoding device for decoding an encoded sequence output from the encoding device according to claim 3,
A cyclic code vector inverse quantization unit that has the reference spectrum sequence and generates the quantized spectrum sequence based on the reference spectrum sequence and a spectrum sequence code included in the encoded sequence;
A Huffman dequantization unit that receives a Huffman code sequence included in the encoded sequence and decodes the Huffman code sequence to generate the quantized spectrum sequence;
A spectrum inverse amplification unit that receives the quantized spectrum sequence and deamplifies the quantized spectrum sequence using a correction gain to generate an amplified spectrum sequence;
A spectral denormalization unit that receives the amplified spectral sequence and converts the amplified spectral sequence using a scale factor to generate a frequency spectral sequence;
As the spectral sequence code included in the encoded data stream is output to the circulating code vector inverse quantization section, a Huffman code string contained in the encoded data stream is output to the Huffman inverse quantization section, the spectral sequence and a code and decoding switching section for switching the Huffman code sequence, decoding device.

Images