JP2005018205A

JP2005018205A - Musical piece retrieving system and method, and musical piece retrieving device

Info

Publication number: JP2005018205A
Application number: JP2003179109A
Authority: JP
Inventors: Takayasu Miki; 孝保三木
Original assignee: Matsushita Electric Industrial Co Ltd
Current assignee: Panasonic Holdings Corp
Priority date: 2003-06-24
Filing date: 2003-06-24
Publication date: 2005-01-20

Abstract

<P>PROBLEM TO BE SOLVED: To provide technology to retrieve a musical piece which is more suitable for the tastes of a user from many musical pieces. <P>SOLUTION: The degree of the tastes of a user for an arbitrary musical piece is inputted by the user himself or herself, or automatically inputted from separately prepared user's taste information. Then, the degree of the tastes of the user for musical pieces including any musical piece unknown to the user are estimated and calculated by using the featured value extracting function, reproduction number of times detecting function, similar musical piece retrieving function and recommendation level calculating function of musical pieces. Thus, a recommended musical piece suitable for the tastes of the user is presented with high precision. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

【０００１】
【発明の属する技術分野】
この発明は、楽曲に対応する特徴量を利用し、楽曲に対する利用者の好みや嗜好の変化に応じて、利用者の好みの楽曲に類似する楽曲を検索し推奨する機能を提供する楽曲検索システム、楽曲検索方法、及び楽曲検索装置に関する。
【０００２】
【従来の技術】
近年、デジタル化された楽曲の販売は、音楽ＣＤやＭＤなどの物理的な媒体を用いるのみならず、インターネット上のサーバにデジタル化し圧縮した楽曲を蓄え、暗号化による著作権保護機能や、課金決済機能を設けたシステムを構築し、利用者が所望する楽曲を、いつでも好きな時間に家庭のパーソナルコンピュータから、インターネット経由で購入することが可能な楽曲販売サービスが普及しつつある。しかし、このような形態でのサービスでは、利用者は、嗜好にあった曲を探すにために、インターネット上の複数の販売サイトを個別にアクセスして、１曲づつ試聴した上で、どの曲を購入すべきかを決定する必要がある。
【０００３】
また、近年、パーソナルコンピュータのハードディスクが大容量化し、また、ハードディスクプレーヤなどの出現で、個人の利用の範囲で、非常に多くの楽曲を蓄積・再生できる環境が提供され、例えば２０ＧＢのハードディスクであれば１曲５ＭＢの楽曲が４０００曲保存できるような環境がひろがりつつあり、利用者自身、自分がどのような楽曲をもっているかを把握しきれない状況が発生しつつある。
【０００４】
また、ジャンルやアーティスト名のような楽曲書誌情報で検索する技術は普及しているが、数千曲の中から利用者の嗜好に応じた適切な楽曲を検索するには十分でない。多くの楽曲の中から、より適切に利用者の嗜好にあった楽曲を検索する技術が望まれている。
【０００５】
一方、音楽の特性を表現する際に使用される音楽表現因子（例えば、明るさや因子や悲しさ因子）を用いて、複数の音楽表現因子ごとに重み付けられた値をもとに楽曲の特性を表現することによって、利用者の嗜好にあった楽曲を検索し販売するサービスが提案されている。しかし、このような形態のシステムを提供するためには、音楽表現因子を用いた分析を各楽曲ごとに実施する必要がある。また、楽曲提供者が試聴システムを構築する場合、楽曲の著作権保護のため、楽曲データと別に試聴データを用意したり、試聴時に楽曲データをコピーされないように保護機能を別途設けるなどの対策を講じる必要があり、システム構築コストが増加する。
【０００６】
また、利用者の再生履歴情報から、利用者の好みの曲を推測する提案を行っているものがある（例えば、特許文献１参照）。しかし、利用者は、いつも好みの曲のみを再生するわけではないので、利用者の嗜好度を推測する精度を確保するのが難しい。
【０００７】
【特許文献１】
特開２００３−９９４５９号公報
【０００８】
【発明が解決しようとする課題】
このように、従来の技術によれば、利用者がインターネット上で好みの曲を探すためには、複数の販売サイトを個別にアクセスする必要があり、かつ、通常データベースに登録されている楽曲名やジャンルのような楽曲の書誌情報からだけでは楽曲の大まかな内容しか推定することができないので、結局、候補曲を１曲づつ試聴しなければならず、利用者にとって未知の楽曲の中から好みの楽曲を探すことは、非常に煩雑な操作になる。
【０００９】
また、個人の利用の範囲で非常に多数の楽曲を有し、その中から自分の好みの楽曲を探す場合、楽曲名やジャンルのような楽曲の書誌情報さえも正しく管理されていない場合もあり、好みの楽曲を探すためには全ての楽曲を再生する必要があるなど、利用者にとって非常に困難な作業になる。
【００１０】
さらに、個人の嗜好は時間によって変化していくものであり、変化する個人嗜好に合わせて適切な楽曲を自動的に選曲する必要があるが、従来のような技術では実現が困難か、あるいは、嗜好度の精度を上げるのが難しい。
【００１１】
【課題を解決するための手段】
かかる課題を解決するための手段として、本発明の請求項１に記載の楽曲検索システムは、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子とを対応付けて管理する特徴量管理手段と、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子とを対応付けて管理する嗜好度管理手段と、楽曲識別子に対応付けられた嗜好度をもとに推奨度１を算出し利用者に対する推奨曲の楽曲識別子を抽出する推奨度算出手段と、推奨曲の楽曲識別子をもとに、特徴量管理手段から推奨曲に対する類似曲の楽曲識別子を抽出する類似曲検索手段とを有し、推奨曲の推奨度１及びあるいは類似曲の類似度から、新たな推奨度２を算出し、利用者に対する推奨曲を算出提示する推奨曲提示手段からなることを特徴とする。
【００１２】
請求項２に記載の楽曲検索システムは、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子とを対応付けて管理する特徴量管理手段と、楽曲に対して利用者が再生した回数と楽曲を識別する楽曲識別子とを対応付けて管理する再生履歴管理手段と、楽曲識別子に対応付けられた再生回数をもとに推奨度１を算出し、利用者に対する推奨曲の楽曲識別子を抽出する推奨度算出手段と、推奨曲の楽曲識別子をもとに、特徴量管理手段から推奨曲に対する類似曲の楽曲識別子を抽出する類似曲検索手段とを有し、推奨曲の推奨度１及びあるいは類似曲の類似度から、新たな推奨度２を算出し、利用者に対する推奨曲を算出提示する推奨曲提示手段からなることを特徴とする。
【００１３】
請求項３に記載の楽曲検索システムは、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子とを対応付けて管理する特徴量管理手段と、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子とを対応付けて管理する嗜好度管理手段と、楽曲に対して利用者が再生した回数と楽曲を識別する楽曲識別子とを対応付けて管理する再生履歴管理手段と、楽曲識別子に対応付けられた嗜好度及びあるいは再生回数をもとに推奨度１を算出し、利用者に対する推奨曲の楽曲識別子を抽出する推奨度算出手段と、推奨曲の楽曲識別子をもとに、特徴量管理手段から推奨曲に対する類似曲の楽曲識別子を抽出する類似曲検索手段とを有し、推奨曲の推奨度１及びあるいは類似曲の類似度から、新たな推奨度２を算出し、利用者に対する推奨曲を算出提示する推奨曲提示手段からなることを特徴とする。
【００１４】
請求項４に記載の楽曲検索システムは、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子とを対応付けて管理する嗜好度管理手段と、楽曲に対して利用者が再生した回数と楽曲を識別する楽曲識別子とを対応付けて管理する再生履歴管理手段と、楽曲識別子に対応付けられた嗜好度及び再生回数をもとに推奨度１を算出し、利用者に対する推奨曲の楽曲識別子を抽出する推奨度算出手段とを有し、推奨曲の推奨度１をもとに、利用者に対する推奨曲を算出提示する推奨曲提示手段からなることを特徴とする。
【００１５】
請求項５に記載の楽曲検索システムは、請求項１から４のいずれか１項に記載の楽曲検索システムにおいて、嗜好度管理手段は、嗜好度として、利用者が楽曲に対して指定する値及びあるいは、好みのアーティストやジャンルなどの楽曲の書誌情報に対する利用者の好みの度合いを所定の係数で嗜好度に積算することを特徴とする。
【００１６】
請求項６に記載の楽曲検索システムは、請求項１から５のいずれか１項に記載の楽曲検索システムにおいて、推奨曲提示手段が推奨曲を算出提示する場合に、該当推奨曲の再生回数が別途定める数値以下、及びあるいは、嗜好度が別途定める数値以下、の楽曲を優先して提示することを特徴とする。
【００１７】
請求項７に記載の楽曲検索システムは、請求項１から６のいずれか１項に記載の楽曲検索システムにおいて、嗜好度管理手段は、利用者の楽曲に対する嗜好の度合いを予めあるいは別途作成されるデータベースから自動的に取得し、楽曲を識別する楽曲識別子と対応付けて管理することを特徴とする。
【００１８】
請求項８に記載の楽曲検索システムは、請求項２から７のいずれか１項に記載の楽曲検索システムにおいて、再生履歴管理手段は、利用者が楽曲を再生した日時及びあるいは一定期間内での再生回数及びあるいは一定期間内再生回数の履歴と、楽曲を識別する楽曲識別子とを対応付けて管理することを特徴とする。
【００１９】
請求項９に記載の楽曲検索システムは、請求項１から８のいずれか１項に記載の楽曲検索システムにおいて、特徴量管理手段、嗜好度管理手段、再生履歴管理手段、推奨度算出手段、類似曲検索手段、推奨曲提示手段の任意の手段が、関連する他の手段との間の処理情報の入出力を、ネットワーク及びあるいはインターネットを経由して処理することを特徴とする。
【００２０】
請求項１０に記載の楽曲検索方法は、利用者に対する推奨曲を検索する楽曲検索方法であって、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子を対応させて管理する特徴量管理データベースと、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子を対応させて管理する嗜好度管理データベースとを有し、嗜好度管理データベースから楽曲の推奨度１を算出し、推奨度算出結果を用いて特徴量管理データベースから楽曲の類似曲とその類似度を抽出し、類似度抽出結果及びあるいは推奨度算出結果を用いて新たな推奨度２を算出し、利用者に対する推奨曲を算出提示することを特徴とする。
【００２１】
請求項１１に記載の楽曲検索方法は、利用者に対する推奨曲を検索する楽曲検索方法であって、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子を対応させて管理する特徴量管理データベースと、楽曲に対して利用者が再生した再生回数と楽曲を識別する楽曲識別子を対応させて管理する再生履歴管理データベースとを有し、嗜好度管理データベースから楽曲の推奨度１を算出し、推奨度算出結果を用いて特徴量管理データベースから楽曲の類似曲とその類似度を抽出し、類似度抽出結果及びあるいは推奨度算出結果を用いて新たな推奨度２を算出し、利用者に対する推奨曲を算出提示することを特徴とする。
【００２２】
請求項１２に記載の楽曲検索方法は、利用者に対する推奨曲を検索する楽曲検索方法であって、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子を対応させて管理する特徴量管理データベースと、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子を対応させて管理する嗜好度管理データベースと、楽曲に対して利用者が再生した再生回数と楽曲を識別する楽曲識別子を対応させて管理する再生履歴管理データベースとを有し、嗜好度管理データベース及びあるいは再生履歴管理データベースから楽曲の推奨度１を算出し、推奨度算出結果を用いて特徴量管理データベースから楽曲の類似曲を抽出し、類似度抽出結果と推奨度算出結果を用いて新たな推奨度２を算出し、推奨度１及びあるいは推奨度２を元に、利用者に対する推奨曲を算出提示することを特徴とする。
【００２３】
請求項１３に記載の楽曲検索方法は、利用者に対する推奨曲を検索する楽曲検索方法であって、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子を対応させて管理する嗜好度管理データベースと、楽曲に対して利用者が再生した再生回数と楽曲を識別する楽曲識別子を対応させて管理する再生履歴管理データベースとを有し、嗜好度管理データベース及び再生履歴管理データベースから楽曲の推奨度１を算出し、推奨曲の推奨度１をもとに、利用者に対する推奨曲を算出提示することを特徴とする。
【００２４】
請求項１４に記載の楽曲検索方法は、請求項１０から１３のいずれか１項に記載の楽曲検索方法において、嗜好度算出時に、利用者が楽曲に対して指定する値及びあるいは、好みのアーティストやジャンルなどの楽曲の書誌情報に対する利用者の好みの度合いを所定の係数で嗜好度算出時に積算することを特徴とする。
【００２５】
請求項１５に記載の楽曲検索方法は、請求項１０から１４のいずれか１項に記載の楽曲検索方法において、推奨曲を算出提示する場合に、該当推奨曲の再生回数が別途定める数値以下、及びあるいは、嗜好度が別途定める数値以下、の楽曲を優先して提示することを特徴とする。
【００２６】
請求項１６に記載の楽曲検索方法は、請求項１０から１５のいずれか１項に記載の楽曲検索方法において、嗜好度算出時に、利用者の楽曲に対する嗜好の度合いを予めあるいは別途作成されるデータベースから自動的に取得し、楽曲を識別する楽曲識別子と対応付けて管理することを特徴とする。
【００２７】
請求項１７に記載の楽曲検索方法は、請求項１１から１６のいずれか１項に記載の楽曲検索方法において、再生履歴管理データベースは、利用者が楽曲を再生した日時及びあるいは一定期間内での再生回数及びあるいは一定期間内再生回数の履歴と、楽曲を識別する楽曲識別子とを対応付けて管理することを特徴とする。
【００２８】
請求項１８に記載の楽曲検索方法は、請求項１０から１７のいずれか１項に記載の楽曲検索方法において、特徴量管理データベース、嗜好度管理データベース、再生履歴管理データベースが、ネットワーク上及びあるいはインターネット上に個別あるいは任意の組み合わせで分散して処理されることを特徴とする。
【００２９】
請求項１９に記載の楽曲検索装置は、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子とを対応付けて管理する特徴量管理手段と、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子と対応つけて管理する嗜好度管理手段と、楽曲識別子に対応付けられた嗜好度をもとに推奨度１を算出し、利用者に対する推奨曲の楽曲識別子を抽出する推奨度算出手段と、推奨曲の楽曲識別子をもとに、特徴量管理手段から推奨曲に対する類似曲の楽曲識別子を抽出する類似曲検索手段とを有し、推奨曲の推奨度１及びあるいは類似曲の類似度から、新たな推奨度２を算出し、利用者に対する推奨曲を算出提示する推奨曲提示手段からなることを特徴とする。
【００３０】
請求項２０に記載の楽曲検索装置は、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子とを対応付けて管理する特徴量管理手段と、楽曲に対して利用者が再生した回数と楽曲を識別する楽曲識別子とを対応付けて管理する再生履歴管理手段と、楽曲識別子に対応付けられた再生回数をもとに推奨度１を算出し、利用者に対する推奨曲の楽曲識別子を抽出する推奨度算出手段と、推奨曲の楽曲識別子をもとに、特徴量管理手段から推奨曲に対する類似曲の楽曲識別子を抽出する類似曲検索手段とを有し、推奨曲の推奨度１及びあるいは類似曲の類似度から、新たな推奨度２を算出し、利用者に対する推奨曲を算出提示する推奨曲提示手段からなることを特徴とする。
【００３１】
請求項２１に記載の楽曲検索装置は、楽曲の特徴を表現する特徴量と楽曲を識別する楽曲識別子とを対応付けて管理する特徴量管理手段と、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子とを対応付けて管理する嗜好度管理手段と、楽曲に対して利用者が再生した回数と楽曲を識別する楽曲識別子とを対応付けて管理する再生履歴管理手段と、楽曲識別子に対応付けられた嗜好度及びあるいは再生回数をもとに推奨度１を算出し、利用者に対する推奨曲の楽曲識別子を抽出する推奨度算出手段と、推奨曲の楽曲識別子をもとに、特徴量管理手段から推奨曲に対する類似曲の楽曲識別子を抽出する類似曲検索手段とを有し、推奨曲の推奨度１及びあるいは類似曲の類似度から、新たな推奨度２を算出し、利用者に対する推奨曲を算出提示する推奨曲提示手段からなることを特徴とする。
【００３２】
請求項２２に記載の楽曲検索装置は、楽曲に対して利用者が指定する嗜好度と楽曲を識別する楽曲識別子とを対応付けて管理する嗜好度管理手段と、楽曲に対して利用者が再生した回数と楽曲を識別する楽曲識別子とを対応付けて管理する再生履歴管理手段と、楽曲識別子に対応付けられた嗜好度及び再生回数をもとに推奨度１を算出し、利用者に対する推奨曲の楽曲識別子を抽出する推奨度算出手段とを有し、推奨曲の推奨度１をもとに、利用者に対する推奨曲を算出提示する推奨曲提示手段からなることを特徴とする。
【００３３】
請求項２３に記載の楽曲検索装置は、請求項１９から２２のいずれか１項に記載の楽曲検索装置において、嗜好度管理手段は、嗜好度として、利用者が楽曲に対して指定する値及びあるいは、好みのアーティストやジャンルなどの楽曲の書誌情報に対する利用者の好みの度合いを所定の係数で嗜好度に積算することを特徴とする。
【００３４】
請求項２４に記載の楽曲検索装置は、請求項１９から２３のいずれか１項に記載の楽曲検索装置において、推奨曲提示手段が推奨曲を算出提示する場合に、該当推奨曲の再生回数が別途定める数値以下、及びあるいは、嗜好度が別途定める数値以下、の楽曲を優先して提示することを特徴とする。
【００３５】
請求項２５に記載の楽曲検索装置は、請求項１９から２４のいずれか１項に記載の楽曲検索装置において、嗜好度管理手段は、利用者の楽曲に対する嗜好の度合いを予めあるいは別途作成されるデータベースから自動的に取得し、楽曲を識別する楽曲識別子と対応付けて管理することを特徴とする。
【００３６】
請求項２６に記載の楽曲検索装置は、請求項２０から２５のいずれか１項に記載の楽曲検索装置において、再生履歴管理手段は、利用者が楽曲を再生した日時及びあるいは一定期間内での再生回数及びあるいは一定期間内再生回数の履歴と、楽曲を識別する楽曲識別子とを対応付けて管理することを特徴とする。
【００３７】
請求項２７に記載の楽曲検索装置は、請求項１９から２６のいずれか１項に記載の楽曲検索装置において、特徴量管理手段、嗜好度管理手段、再生履歴管理手段、推奨度算出手段、類似曲検索手段、推奨曲提示手段の任意の手段が、関連する他の手段との間の処理情報の入出力を、ネットワーク及びあるいはインターネットを経由して処理することを特徴とする。
【００３８】
請求項２８及び請求項２９に記載のコンピュータに実行させるためのプログラム、及び前記プログラムを記録したコンピュータ読み取り可能な記録媒体は、請求項１から９のいずれか１項に記載の楽曲検索システムの動作、あるいは、請求項１０から１８のいずれか１項に記載の楽曲検索方法により実現される機能、あるいは、請求項１９から２７のいずれか１項に記載の楽曲検索装置の動作を実行させることを特徴とする。
【００３９】
【発明の実施の形態】
以下、この発明の実施の形態について図面を参照して説明する。図１はこの発明が適用された楽曲検索システムの構成の一例を示すものである。特徴量抽出手段１０は、楽曲の音響信号から予め定めた特徴量を抽出し、特徴量管理手段１１は、抽出された前記特徴量を、前記楽曲を識別する楽曲識別子と対応させて管理する。嗜好度入力手段１２は、利用者の特定の楽曲に対する嗜好度を利用者自身が指定するための入力手段であり、嗜好度管理手段１３は、利用者が入力した前記嗜好度を、前記楽曲を識別する楽曲識別子と対応させて管理する。再生検出手段１４は、利用者が楽曲を再生した場合に、再生履歴管理手段１５に再生検出を伝え、再生履歴管理手段１５は、再生された前記楽曲を識別する楽曲識別子と対応してもつ再生カウントを追加更新し保管管理する。推奨度算出手段１６は、嗜好度管理手段１３及びあるいは再生履歴管理手段１５にて楽曲識別子と対応させて管理されている楽曲嗜好度及びあるいは楽曲再生履歴をもとに、利用者に推奨すべき楽曲の１次推奨曲情報として楽曲識別子とその１次推奨度ＦＣを対応させて算出する。類似曲検索手段１７は、推奨度算出手段１６で算出された１次推奨曲情報の楽曲識別子を入力し、特徴量管理手段１１にで管理される楽曲の特徴量をもとに、前記楽曲識別子に対応する楽曲の類似曲を検索し、検索結果の楽曲識別子及び類似度Ｒを推奨曲提示手段１８に出力する。推奨曲提示手段１８は、推奨度算出手段１６の１次算出結果である１次推奨曲情報及びあるいは、類似曲検索手段１７の出力である前記検索結果の楽曲識別子及び類似度Ｒ、を用いて利用者に推奨すべき２次推奨曲情報として楽曲識別子と２次推奨度ＲＥＣを対応させて算出し、その結果の中から別途指定する基準に適合するものを利用者に推奨曲として提示する。
【００４０】
以下、各部の動作の詳細について、図２のフローチャートに基づき説明する。利用者は、特徴量抽出手段１０に対し、自分の所有する楽曲を指定する。ステップＳ１０では、指定された楽曲の音響信号から、予め定められた特徴量を抽出する。特徴量抽出手段１０で抽出される特徴量として、スペクトル変化度Ｐ１（フレーム間のスペクトル変化の度合い）、平均発音数Ｐ２（楽曲中で発音される音の発音頻度）、発音非周期性Ｐ３（楽曲中で発音される音の非周期性の度合い）、拍周期Ｐ４（楽曲の４分音符に相当する時間長）、が上げられるが、上記４種類の特徴量の他に、拍周期比率Ｐ５、拍強度Ｐ６、拍強度比Ｐ７などのパラメータについても特徴量として利用しても良い。なお、対象となる楽曲から特徴量を抽出するにあたり、特徴量の抽出範囲は、楽曲の全体、楽曲の一部分、及び楽曲の複数の部分に対し、任意の組み合わせからなる領域から抽出して良い。また、ステップＳ１０における、別の特徴量の求め方として、ＳＤ（ｓｅｍａｎｔｉｃｄｉｆｆｅｒｅｎｃｉａｌ）法等による官能評価実験などにより、楽曲に対する利用者の主観的な印象をその楽曲に対する複数の特徴量として予め求めて楽曲識別子と対応させてデータベース化しておき、楽曲識別子をキーとして、指定された楽曲に対応する特徴量を検索して求める方法でも良い。ステップＳ１１では、前述のステップＳ１０で求めた特徴量を該当楽曲の楽曲識別子と対応させて、特徴量管理手段１１にて特徴量ＤＢ２１に格納管理する。特徴量ＤＢ２１の内容の例を、図３、特徴量データ３１に示す。特徴量ＤＢ２１は、楽曲情報３２のフォーマットで楽曲識別子と楽曲の歌手名や楽曲所在場所に関する情報を同時に管理するようにしても良い。
【００４１】
一方、ステップＳ１２では再生検出手段１４により利用者が楽曲を再生したことを検出し、再生履歴管理手段１５に通知する。ステップＳ１３では通知に従い再生履歴管理手段１５にて再生回数ＤＢ２２にアクセスし、該当曲の楽曲識別子に対応する再生カウントを１加算する。この場合に加算する量は、別途定めた係数によって算出しても良い。さらに、ステップＳ１４では、嗜好度入力手段１２により、利用者に対し、利用者自身が所定の楽曲に対する自分の好み度合いである嗜好度Ｆを入力させる。本実施の形態では、嗜好度Ｆは０または正の値をとることを前提とするが、値の範囲を正負にすることにより好き（正）嫌い（負）を表現することにしても良い。また、前記嗜好度入力手段１２では、利用者自身が嗜好度Ｆを入力するようにしたが、利用者の指定された楽曲に対する嗜好の度合いが、予め別途保管されておりそのデータが利用できる場合には、適切な処理をもって自動入力させるようにしても良い。ステップＳ１５では、入力された嗜好度Ｆを楽曲識別子と対応させて嗜好度管理手段１３により嗜好度ＤＢ２３に書き込む。再生回数ＤＢ２２と嗜好度ＤＢ２３は、図３にある個人嗜好情報３３のフォーマットを使用して単一のデータベースとして管理しても良い。
【００４２】
さて、利用者から推奨曲提示を依頼された場合には、ステップＳ１６により各楽曲の嗜好度及びあるいは再生回数を抽出する。これは、推奨度算出手段１６により、再生履歴管理手段１５が管理する再生回数ＤＢ２２内のデータから楽曲識別子ｎに対応する再生回数Ｃｎ及びあるいは、嗜好度管理手段１５が管理する嗜好度ＤＢ２３内のデータから楽曲識別子ｎに対応する嗜好度Ｆｎを抽出することによって行う。
【００４３】
ステップＳ１７では、推奨度算出手段１６により、１次推奨度ＦＣｎを各ＤＢ内に登録された楽曲ごとに算出する。
【００４４】
【数１】

【００４５】
（ａ１，ａ２は、１次推奨度ＦＣｎに対する、嗜好度と再生回数の影響度を調整するための係数）
さらに、ＦＣｎの値の大きいものから順に別途定める曲数ｙだけ、それぞれのＦＣｎとそれに対応する楽曲識別子ｎを対にして抽出し、類似曲検索手段１７に出力する。このとき出力される楽曲識別子ｎと１次推奨度ＦＣｎを対にした要素の集合を１次推奨度配列｛ＦＣＸ｝とする。１次推奨度配列｛ＦＣＸ｝の要素数はｙになる。
【００４６】
ステップＳ１８では、類似曲検索手段１７により、１次推奨度配列｛ＦＣＸ｝の要素である各楽曲識別子ｎのそれぞれに対する類似曲の楽曲識別子を検索する。
【００４７】
次に、類似曲の検索方法を説明する。１次推奨度配列｛ＦＣＸ｝の要素について順にそれぞれの類似曲を求める。ある要素（楽曲識別子ｎ）の特徴量Ｐｎ（Ｐｎ１、Ｐｎ２、Ｐｎ３、…ＰｎＮ）が与えられた場合、特徴量ＤＢ２１で保管されている各楽曲の楽曲識別子ｍの特徴量Ｐｍ（Ｐｍ１、Ｐｍ２、Ｐｍ３…ＰｍＮ）と、検索キーとして与えられた楽曲識別子ｎの特徴量Ｐｎ（Ｐｎ１、Ｐｎ２、Ｐｎ３、…ＰｎＮ）との距離を、数２にて各特徴量間のユークリッド距離として算出し、その逆数を、類似度を表す値Ｒｎｍとして順次求める。
【００４８】
【数２】

【００４９】
（Ｎは本実施の形態で使用される特徴量の数、Ｂｉは各特徴量を正規化するための係数。ｎは１次推奨度配列｛ＦＣＸ｝のｎ番目の要素、ｍは特徴量ＤＢ２１で保管されている楽曲のｍ番目の要素に対応することを表す。）
ここでは、Ｒｎｍの算出は、ｎについては、１次推奨度配列｛ＦＣＸ｝のすべての要素の楽曲を算出対象とし、ｍについては、特徴量ＤＢ２１で保管されているすべての楽曲を算出対象として、類似度を表す値Ｒｎｍを算出することとする。
【００５０】
次に、ステップＳ１９では推奨曲提示手段１８により、以下のような方法で２次推奨度ＲＥＣｎｘを算出する。
【００５１】
まず、要素数ｙの１次推奨度配列配列｛ＦＣＸ｝のｎ番目の要素である楽曲識別子ｎに対する類似曲を求める。これは、楽曲識別子ｎの類似曲である楽曲識別子ｍとその類似度Ｒｎｍを対応させて配列の要素とし、Ｒｎｍの大きいもの順に別途定めた曲数ｚだけ整列させた配列｛ＴＥＭＰｎ｝として作成する。配列｛ＴＥＭＰｎ｝の要素数はｚになる。ここで、配列｛ＴＥＭＰｎ｝は１次推奨度配列配列｛ＦＣＸ｝の要素ごとに作成される。また、配列｛ＴＥＭＰｎ｝のｘ番目にある要素を、楽曲識別子ｎｘと類似度Ｒｎｘと呼ぶことにする。楽曲識別子ｎと楽曲識別子ｎｘとの類似度Ｒｎｘは各特徴量間のユークリッド距離が非常に近ければ極めて大きな値になるため、ここでは１例として、類似度Ｒｎｘを変換して近似度Ｓｎｘを数３によって求める。近似度Ｓｎｘは１〜０の値の間で変化する楽曲間の近似度を表す数値である。
【００５２】
【数３】

【００５３】
（ｄ１、ｄ２は、近似度を調整する定数。本例ではｄ１、ｄ２は正の値とする）
楽曲識別子ｎの１次推奨度はＦＣｎとして既に求められているので、楽曲識別子ｎｘの２次推奨度ＲＥＣｎｘは、以下のように数４によって求める。
【００５４】
【数４】

【００５５】
次に、配列｛ＴＥＭＰｎ｝の要素である、楽曲識別子ｎｘと類似度Ｒｎｘの対のうち類似度Ｒｎｘを２次推奨度ＲＥＣｎｘに置き換えた配列を配列｛ＴＥＭＰＲｎ｝とする。配列｛ＴＥＭＰＲｎ｝はｚ個の要素をもつ。
【００５６】
その後、配列｛ＦＣＸ｝のｙ個の各要素から算出される、要素数ｚの配列｛ＴＥＭＰｎ｝を順次ｙ個算出し、その結果を２次元の（ｙ、ｚ）配列である２次推奨度配列｛ＲＥＣｎｘ｝とする。２次推奨度配列｛ＲＥＣｎｘ｝は、２次推奨度ＲＥＣｎｘとそれに対応する楽曲識別子ｊを対にした要素からなる２次元配列である。
【００５７】
ステップＳ２０では、推奨曲提示手段１８により算出された２次推奨度配列｛ＲＥＣｎｘ｝の各要素の楽曲識別子ｊをもとに、その楽曲名などの書誌情報を取得し、２次推奨度ＲＥＣｎｘの大きいもの順に、利用者に提示する。
【００５８】
なお、以上は、２次推奨度ＲＥＣｎｘ及び２次推奨度配列｛ＲＥＣｎｘ｝の算出の１例であり、前記配列｛ＴＥＭＰＲｎ｝の先頭に、配列｛ＴＥＭＰＲｎ｝の算出の元にした配列｛ＦＣＸ｝の要素である楽曲識別子ｎと、前記楽曲識別子ｎに対応する２次推奨度の値として２次推奨度ＲＥＣｎｘの代わりにＦＣｎを対応させた要素を１つ追加し、配列｛ＴＥＭＰＲｎ｝の要素数を（ｚ＋１）として２次推奨度配列｛ＲＥＣｎｘ｝をもとめても良い。
【００５９】
また、配列の要素数ｙ、ｚを定数としたが、各配列算出時に配列ごとに可変数にしても良い。また、１次推奨度ＦＣｎは、各楽曲に対する嗜好度Ｆと再生回数Ｃから算出したが、別途、利用者に好みのアーティストやジャンルなどの楽曲書誌的項目に対する嗜好度を入力させ、各楽曲の嗜好度に反映させても良い。この場合は、好みのジャンルの嗜好度が入力されれば、そのジャンルに属する楽曲に対する嗜好度を係数ｅ１によって自動的に加算するような仕組みにすることが考えられる。また、推奨曲提示手段１８で利用者に提示される推奨曲は、上記実施例の２次推奨度配列｛ＲＥＣｎｘ｝から抽出しても良いし、１次推奨度配列｛ＦＣＸ｝から抽出しても良い。また、嗜好度及びあるいは再生回数等から算出する１次推奨度配列｛ＦＣＸ｝の要素である各楽曲のうち、再生回数が別途定める定数ｆ１以下のものを特に優先して抽出するようにしても良い。このことにより、利用者は好みの楽曲に類似する曲の中から、未だ試聴回数が少ないもの、あるいは未だ試聴したことがないものから優先的に提示を受けることが可能になる。
【００６０】
また、本実施の形態では、嗜好度Ｆは０または正の値をとることを前提としたが、値の範囲を正負にすることにより、好き（正）嫌い（負）を表現することにしても良い。このことにより、以前は嫌いの分類にいれていた楽曲に対しても、試しに（気分転換として）試聴するための推奨曲を提示させることが可能になる。
【００６１】
また、楽曲に対する嗜好度を、利用者がそのときの気分によって変更できるようにしたり、楽曲の再生回数だけでなく直近の一定期間の再生回数や一定期間再生回数の時間変化を１次推奨度ＦＣｎ算出時のパラメータとして追加することにより、２次推奨度配列｛ＲＥＣｎｘ｝や１次推奨度配列｛ＦＣＸ｝で算出される推奨曲を、変化する利用者の気分に合わせてより適切な推奨結果を算出するようにしても良い。さらに、上記実施例中の数１の係数ａ１，ａ２や、数３の係数ｄ１，ｄ２、あるいは係数ｅ１，ｆ１を調節できるようにすることにより、より利用者の好みに適合する推奨曲の算出を可能にしても良い。
【００６２】
さらに、上記の実施例は、単一のパーソナルコンピュータや、オーディオプレーヤなどの機器で完結した機能として実施しても良いし、それらがネットワークで接続された環境で、図１の各手段１０〜１８を分散したり一部共有した任意の形態で実施してもよく、さらに図４に示すように、インターネット上の楽曲サービスプロバイダー４１に特徴量楽曲識別子及び楽曲コンテンツデータベース４２を持たせ、パーソナルコンピュータなどの利用者端末４３やオーディオプレーヤ４３１では、楽曲識別子嗜好度再生回数データベース４４を持たせて実施しても良いし、各構成要素を別の形態、すなわち、インターネット上で、図１の各手段１０〜１８を分散したり一部共有した任意の形態で実施しても良い。
【００６３】
また、楽曲サービスプロバイダー４１が楽曲データを利用者に送付・配信する際に、楽曲データに対応する特徴量及びあるいは楽曲情報を、楽曲の楽曲データと共にあるいは別途関連させて、図４の構成要素である利用者端末４３に送付するようにし、利用者端末４３やオーディオプレーヤ４３１からは、利用者の楽曲に対する嗜好度や再生回数を楽曲サービスプロバイダー４１に随時返すような形態にすることにより、楽曲サービスプロバイダー４１が、利用者の嗜好を反映した販売曲の推奨を可能にできるような形態にしても良い。
【００６４】
また、楽曲サービスプロバイダー４１は、既存のコンテンツプロバイダーと提携し、少なくとも推奨曲提示手段１８の機能を利用者に提供することによって、利用者に対して検索使用料や、コンテンツプロバイダーに対して楽曲紹介料・販売促進料などを請求し、収入とするような形態にしても良い。また、前記形態において、少なくとも嗜好度算出手段１２及びあるいは再生検出手段１４を、ＰＣなどの利用者端末側で実行する機能を利用者に有償または無償で提供することにより、コンテンツプロバイダーなどの第３者から利用度合いに応じたライセンス料を徴収する形態にしても良い。
【００６５】
【発明の効果】
本発明により、利用者が新たに入手を希望する楽曲の検索時や、あるいは膨大な曲を所有して利用者自身が自分の所有する楽曲を把握しきれなくなったような状況において、利用者の好みにあった楽曲をより高い精度で推奨提示することが可能になる。本発明は、利用者が、初期あるいは利用に応じて一定数の楽曲に対する好みの度合いを嗜好度として入力するだけで、その他は、自動的に楽曲の特徴量や再生回数が算出され、類似曲検索機能や推奨度算出機能により、利用者の好みにあった曲を精度高く抽出提示可能になる。また、長く使えば使うほど個人嗜好度や再生回数が個人嗜好情報として蓄積され、より精度高く抽出提示できると同時に、利用者の好みの変化も個人嗜好情報に反映されていくため、利用者の嗜好の変化に応じた推奨曲の提示が可能になる。また、膨大な曲数の中から、利用者の好みに類似した曲でありながら、利用者がまだ１度も試聴したことがないような楽曲を優先して提示することも可能になる。
【００６６】
また、検索サービス提供者は、既存のコンテンツデータ提供者が所有する楽曲の特徴量を予め抽出し、楽曲ＩＤなど他の情報と共に登録する作業を行うだけで、コンテンツデータ提供者の提供する楽曲を利用者の検索対象にすることができ、利用者の個人嗜好情報を利用することによって、利用者の嗜好や嗜好の変化に応じた、上記のようなさまざまな機能による推奨曲の提示が可能になる。
【図面の簡単な説明】
【図１】本発明の実施の形態による楽曲検索システムの概略構成を表すブロック図
【図２】本発明の実施の形態による楽曲検索システムの動作の説明に用いるフローチャート
【図３】本発明の実施の形態による楽曲検索システムで扱うレコードの例を示す図
【図４】本発明の実施の形態による楽曲検索システムの１例の概要を表すシステム図
【符号の説明】
１０特徴量抽出手段
１１特徴量管理手段
１２嗜好度入力手段
１３嗜好度管理手段
１４再生検出手段
１５再生履歴管理手段
１６推奨度算出手段
１７類似曲検索手段
１８推奨曲提示手段
２１特徴量ＤＢ
２２再生回数ＤＢ
２３嗜好度ＤＢ
３１特徴量データ
３２楽曲情報
３３個人嗜好度情報
４１楽曲サービスプロバイダー
４２特徴量楽曲識別子及び楽曲コンテンツデータベース
４３利用者端末
４３１オーディオプレーヤ
４４楽曲識別子嗜好度再生回数データベース
４５インターネット[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a music search system that uses a feature amount corresponding to a music and provides a function for searching for and recommending music similar to a user's favorite music according to a user's preference or preference change with respect to the music. The present invention relates to a music search method and a music search device.
[0002]
[Prior art]
In recent years, the sale of digitized music not only uses physical media such as music CDs and MDs, but also stores music that has been digitized and compressed on a server on the Internet, copyright protection function by encryption, charging Music sales services that allow users to purchase a music composition desired by a user from a personal computer at home at any time, via the Internet, have become popular. However, in this form of service, in order to search for a song that suits the user's preference, the user accesses a plurality of sales sites on the Internet individually, listens to each song, and selects which song. Need to decide what to buy.
[0003]
In recent years, the capacity of personal computer hard disks has increased, and with the advent of hard disk players and the like, an environment has been provided in which a large amount of music can be stored and played within the range of personal use. For example, a 20 GB hard disk can be used. For example, an environment in which 4000 pieces of music of 5 MB can be stored is spreading, and a situation in which the user himself cannot grasp what kind of music he / she has is occurring.
[0004]
In addition, techniques for searching by music bibliographic information such as genre and artist name are widespread, but it is not sufficient to search for appropriate music according to user's preference from among thousands of songs. There is a demand for a technique for searching for music that suits the user's taste more appropriately from many music.
[0005]
On the other hand, using music expression factors (for example, brightness, factors, and sadness factors) used to express the characteristics of music, the characteristics of the music can be determined based on values weighted for each of the music expression factors. A service for searching and selling music that suits the user's preference by expressing it has been proposed. However, in order to provide such a system, it is necessary to perform analysis using music expression factors for each piece of music. In addition, when a music provider builds a trial listening system, in order to protect the copyright of the music, take measures such as preparing trial listening data separately from the music data, or providing a separate protection function so that the music data is not copied during the trial listening. It is necessary to take this, and the system construction cost increases.
[0006]
In addition, there is a proposal that estimates a user's favorite music from user's reproduction history information (see, for example, Patent Document 1). However, since the user does not always play only the favorite song, it is difficult to ensure the accuracy of estimating the user's preference.
[0007]
[Patent Document 1]
JP 2003-99459 A
[0008]
[Problems to be solved by the invention]
As described above, according to the conventional technology, in order for a user to search for a favorite song on the Internet, it is necessary to individually access a plurality of sales sites, and a song name registered in a normal database is used. Since only the general content of a song can be estimated from the bibliographic information of the song, such as the genre and the genre, eventually, the candidate song must be auditioned one by one, and the user likes the song that is unknown to the user. Searching for a song is a very complicated operation.
[0009]
Also, if you have a very large number of songs for personal use and you search for your favorite song from among them, even the bibliographic information of the song, such as song name and genre, may not be managed correctly In order to search for a favorite song, it is necessary to reproduce all the songs, which is very difficult for the user.
[0010]
In addition, personal preferences change with time, and it is necessary to automatically select appropriate music according to the changing personal preferences, but it is difficult to achieve with conventional techniques, It is difficult to improve the accuracy of preference.
[0011]
[Means for Solving the Problems]
As a means for solving such a problem, the music search system according to claim 1 of the present invention is a feature quantity management means for managing a feature quantity expressing a feature of a song in association with a song identifier for identifying the song. And a preference level management means for managing a preference level designated by the user for a song and a song identifier for identifying the song, and a recommendation level of 1 based on the preference level associated with the song identifier. A degree-of-recommendation calculating means for calculating and extracting a song identifier of a recommended song for a user, and a similar song searching means for extracting a song identifier of a similar song to the recommended song from the feature amount management means based on the song identifier of the recommended song And a recommended music presenting means for calculating a new recommended degree 2 from the recommended degree 1 of the recommended music and / or the similarity of the similar music, and calculating and presenting the recommended music to the user.
[0012]
The music search system according to claim 2 is characterized in that a feature quantity management unit that manages a feature quantity that expresses a feature of a song and a song identifier that identifies the song in association with each other, and a number of times the user has played the song. Based on the reproduction history management means for managing the music identifier for identifying the music in association with each other, the recommended degree 1 is calculated based on the number of times of reproduction associated with the music identifier, and the music identifier of the recommended music for the user is extracted. Recommended degree calculation means, and similar music search means for extracting music identifiers of similar songs to the recommended music from the feature quantity management means based on the music identifier of the recommended music, and the recommended music recommendation degree 1 and / or similar It comprises a recommended song presentation means for calculating a new recommendation degree 2 from the similarity of songs and calculating and presenting a recommended song for the user.
[0013]
The music search system according to claim 3, a feature amount management unit that manages a feature amount that expresses a feature of a song and a song identifier that identifies the song in association with each other, and a preference level designated by the user for the song A preference level management means for managing the music identifier for identifying the music in association with each other, a playback history management means for managing the number of times the user has played the music and the music identifier for identifying the music in association with each other, Based on the preference level calculating means for calculating the recommendation level 1 based on the preference level and / or the number of times of reproduction associated with the music identifier and extracting the music identifier of the recommended music for the user, and the music identifier of the recommended music And a similar song search means for extracting a song identifier of a similar song with respect to the recommended song from the feature amount management means, and a new recommendation degree 2 is calculated from the recommendation degree 1 of the recommended song and / or the similarity degree of the similar song, Recommendations for users Characterized by comprising the recommended music presented means for calculating presenting songs.
[0014]
The music search system according to claim 4 is a preference level management means for managing a preference level specified by a user for a song and a song identifier for identifying the song in association with each other; A reproduction history management means that associates and manages the number of times played and a music identifier that identifies the music, and calculates a recommended degree 1 based on the preference level and the number of times of reproduction that are associated with the music identifier, and recommends music to the user And a recommended music calculation means for extracting and recommending music based on a recommended music recommendation level of 1.
[0015]
The music search system according to claim 5 is the music search system according to any one of claims 1 to 4, wherein the preference level management means includes a value designated by the user for the music as a preference level, and Alternatively, the degree of preference of the user with respect to the bibliographic information of music such as a favorite artist or genre is integrated with the degree of preference by a predetermined coefficient.
[0016]
A music search system according to a sixth aspect is the music search system according to any one of the first to fifth aspects, wherein when the recommended music presenting means calculates and presents the recommended music, the number of times the corresponding recommended music is reproduced is displayed. It is characterized by preferentially presenting music pieces that have a numerical value that is separately determined and / or that have a preference level that is separately specified.
[0017]
The music search system according to claim 7 is the music search system according to any one of claims 1 to 6, wherein the preference level management means creates a degree of preference of the user's music in advance or separately. It is automatically acquired from a database and managed in association with a music identifier for identifying music.
[0018]
The music search system according to claim 8 is the music search system according to any one of claims 2 to 7, wherein the reproduction history management means includes a date and / or time within a predetermined period when the user reproduces the music. A history of the number of times of reproduction and / or the number of times of reproduction within a certain period and a music identifier for identifying music are associated with each other and managed.
[0019]
The music search system according to claim 9 is the music search system according to any one of claims 1 to 8, wherein the feature amount management unit, the preference level management unit, the reproduction history management unit, the recommendation level calculation unit, and the like Arbitrary means of music search means and recommended music presentation means process input / output of processing information with other related means via a network and / or the Internet.
[0020]
The music search method according to claim 10 is a music search method for searching for a recommended song for a user, wherein a feature value that manages a feature value that expresses a feature of a song and a song identifier that identifies the song are associated with each other. A database and a preference level management database for managing a preference level designated by a user for a song and a song identifier for identifying the song, and calculating a recommendation level 1 of the song from the preference level management database; Use the recommendation level calculation result to extract similar songs and their similarities from the feature management database, calculate the new recommendation level 2 using the similarity extraction result and / or the recommendation level calculation result, and recommend to users The music is calculated and presented.
[0021]
The music search method according to claim 11 is a music search method for searching for a recommended music for a user, wherein a feature value that manages a feature value that expresses a feature of a song and a song identifier that identifies the song are associated with each other. A database and a playback history management database that manages the number of playbacks played by a user for a song and a song identifier that identifies the song, and calculates a recommendation degree 1 of the song from the preference management database; Use the recommendation level calculation result to extract similar songs and their similarities from the feature management database, calculate the new recommendation level 2 using the similarity extraction result and / or the recommendation level calculation result, and recommend to users The music is calculated and presented.
[0022]
The music search method according to claim 12 is a music search method for searching for a recommended song for a user, wherein a feature value that manages a feature value that expresses a feature of a song and a song identifier that identifies the song are associated with each other. A database, a preference management database that manages a preference specified by a user for a song and a song identifier that identifies the song, and a song that identifies the number of playbacks and a song played by the user for the song A reproduction history management database that manages the identifiers in association with each other, calculates a recommendation degree 1 of the music from the preference management database and / or the reproduction history management database, and uses the recommendation degree calculation result to calculate the music Similar music is extracted, a new recommendation level 2 is calculated using the similarity extraction result and the recommendation level calculation result, and the user is calculated based on the recommendation level 1 and / or the recommendation level 2. And calculates presented a recommended song against.
[0023]
The music search method according to claim 13 is a music search method for searching for music recommended to the user, and manages the music specified by the user and the music identifier for identifying the music in association with each other. It has a preference management database and a playback history management database that manages the number of playbacks played by a user for a song and a song identifier that identifies the song, and tracks music from the preference management database and the playback history management database. The recommended degree 1 is calculated, and the recommended music for the user is calculated and presented based on the recommended degree 1 of the recommended music.
[0024]
The music search method according to claim 14 is the music search method according to any one of claims 10 to 13, wherein a value designated by the user for the music and / or a favorite artist when calculating the preference level. The degree of preference of the user with respect to the bibliographic information of the music such as the genre or the genre is integrated when calculating the degree of preference by a predetermined coefficient.
[0025]
The music search method according to claim 15 is the music search method according to any one of claims 10 to 14, wherein when the recommended music is calculated and presented, the number of times of reproduction of the corresponding recommended music is equal to or less than a numerical value separately defined, Alternatively, music having a preference degree equal to or less than a numerical value separately determined is preferentially presented.
[0026]
The music search method according to claim 16 is the music search method according to any one of claims 10 to 15, wherein the preference degree of the user's music is created in advance or separately when calculating the preference level. Are automatically acquired from and managed in association with a music identifier for identifying music.
[0027]
The music search method according to claim 17 is the music search method according to any one of claims 11 to 16, wherein the reproduction history management database stores the date and / or time within a predetermined period when the user reproduced the music. A history of the number of times of reproduction and / or the number of times of reproduction within a certain period and a music identifier for identifying music are associated with each other and managed.
[0028]
The music search method according to claim 18 is the music search method according to any one of claims 10 to 17, wherein the feature quantity management database, the preference management database, and the reproduction history management database are on a network and / or the Internet. It is characterized by being processed individually or in an arbitrary combination.
[0029]
The music search apparatus according to claim 19, a feature amount management unit that manages a feature amount that expresses a feature of a song and a song identifier that identifies the song in association with each other, and a preference level designated by the user for the song And a preference level management means for managing the song in association with the song identifier, and calculating a recommendation degree 1 based on the preference level associated with the song identifier, and extracting the song identifier of the recommended song for the user Recommended degree calculation means, and similar music search means for extracting music identifiers of similar songs to the recommended music from the feature quantity management means based on the music identifier of the recommended music, and the recommended music recommendation degree 1 and / or similar It comprises a recommended song presentation means for calculating a new recommendation degree 2 from the similarity of songs and calculating and presenting a recommended song for the user.
[0030]
The music search device according to claim 20 is characterized in that a feature quantity managing means for managing a feature quantity that expresses a feature of a song and a song identifier that identifies the song in association with each other, and the number of times the user has played the song. Based on the reproduction history management means for managing the music identifier for identifying the music in association with each other, the recommended degree 1 is calculated based on the number of times of reproduction associated with the music identifier, and the music identifier of the recommended music for the user is extracted. Recommended degree calculation means, and similar music search means for extracting music identifiers of similar songs to the recommended music from the feature quantity management means based on the music identifier of the recommended music, and the recommended music recommendation degree 1 and / or similar It comprises a recommended song presentation means for calculating a new recommendation degree 2 from the similarity of songs and calculating and presenting a recommended song for the user.
[0031]
The music search apparatus according to claim 21, wherein a feature amount management unit that manages a feature amount that expresses a feature of a song and a song identifier that identifies the song in association with each other, and a preference level designated by the user for the song A preference level management means for managing the music identifier for identifying the music in association with each other, a playback history management means for managing the number of times the user has played the music and the music identifier for identifying the music in association with each other, Based on the preference level calculating means for calculating the recommendation level 1 based on the preference level and / or the number of times of reproduction associated with the music identifier and extracting the music identifier of the recommended music for the user, and the music identifier of the recommended music And a similar song search means for extracting a song identifier of a similar song with respect to the recommended song from the feature amount management means, and a new recommendation degree 2 is calculated from the recommendation degree 1 of the recommended song and / or the similarity degree of the similar song, Recommendations for users Characterized in that it consists of recommended songs presented means for calculating presented.
[0032]
The music search device according to claim 22 is a preference level management means for managing a preference level specified by a user for a song and a song identifier for identifying the song in association with each other; A reproduction history management means that associates and manages the number of times played and a music identifier that identifies the music, and calculates a recommended degree 1 based on the preference level and the number of times of reproduction that are associated with the music identifier, and recommends music to the user And a recommended music calculation means for extracting and recommending music based on a recommended music recommendation level of 1.
[0033]
The music search device according to claim 23 is the music search device according to any one of claims 19 to 22, wherein the preference level management means includes a value designated by the user for the music as a preference level, and Alternatively, the degree of preference of the user with respect to the bibliographic information of music such as a favorite artist or genre is integrated with the degree of preference by a predetermined coefficient.
[0034]
The music search device according to claim 24 is the music search device according to any one of claims 19 to 23, wherein when the recommended song presentation means calculates and presents the recommended song, the number of times the corresponding recommended song is played back is calculated. It is characterized by preferentially presenting music pieces that have a numerical value that is separately determined and / or that have a preference level that is separately specified.
[0035]
The music search device according to claim 25 is the music search device according to any one of claims 19 to 24, wherein the preference level management means creates the degree of preference of the user's music in advance or separately. It is automatically acquired from a database and managed in association with a music identifier for identifying music.
[0036]
The music search device according to claim 26 is the music search device according to any one of claims 20 to 25, wherein the reproduction history management means includes the date and time when the user reproduces the music and / or within a certain period of time. A history of the number of times of reproduction and / or the number of times of reproduction within a certain period and a music identifier for identifying music are associated with each other and managed.
[0037]
The music search device according to claim 27 is the music search device according to any one of claims 19 to 26, wherein the feature amount management means, the preference degree management means, the reproduction history management means, the recommendation degree calculation means, and the like Arbitrary means of music search means and recommended music presentation means process input / output of processing information with other related means via a network and / or the Internet.
[0038]
A program for causing a computer to execute according to claim 28 and a computer-readable recording medium on which the program is recorded is an operation of the music search system according to any one of claims 1 to 9. Or a function realized by the music search method according to any one of claims 10 to 18, or an operation of the music search device according to any one of claims 19 to 27. Features.
[0039]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an example of the configuration of a music search system to which the present invention is applied. The feature quantity extraction means 10 extracts a predetermined feature quantity from the acoustic signal of the music, and the feature quantity management means 11 manages the extracted feature quantity in association with the music identifier for identifying the music. The preference level input means 12 is an input means for the user himself / herself to specify the preference level for the user's specific music, and the preference level management means 13 uses the preference level input by the user as the music level. It is managed in correspondence with the music identifier to be identified. When the user reproduces the music, the reproduction detection means 14 notifies the reproduction history management means 15 of the reproduction detection, and the reproduction history management means 15 reproduces the music corresponding to the music identifier for identifying the reproduced music. Add and update counts and manage them. The recommendation level calculation means 16 should be recommended to the user on the basis of the music preference level and / or the music playback history managed in association with the music identifier by the preference level management means 13 and / or the playback history management means 15. As the primary recommended music information of the music, the music identifier and the primary recommendation FC are calculated in association with each other. The similar music search means 17 inputs the music identifier of the primary recommended music information calculated by the recommendation level calculation means 16, and based on the feature quantity of the music managed by the feature quantity management means 11, the music identifier The similar music of the music corresponding to is searched, and the music identifier and similarity R of the search result are output to the recommended music presentation means 18. The recommended song presentation unit 18 uses the primary recommended song information that is the primary calculation result of the recommendation degree calculation unit 16 and the song identifier and the similarity R of the search result that is the output of the similar song search unit 17. As the secondary recommended music information to be recommended to the user, the music identifier and the secondary recommendation level REC are calculated in association with each other, and those that meet the criteria specified separately from the results are presented to the user as recommended music.
[0040]
Hereinafter, the details of the operation of each unit will be described based on the flowchart of FIG. The user designates the musical piece owned by the user with respect to the feature amount extraction means 10. In step S10, a predetermined feature amount is extracted from the acoustic signal of the designated music piece. As feature quantities extracted by the feature quantity extraction means 10, the degree of spectrum change P1 (the degree of spectrum change between frames), the average number of pronunciations P2 (the pronunciation frequency of sounds produced in music), the pronunciation aperiodicity P3 ( The degree of non-periodicity of the sound produced in the music) and the beat period P4 (the length of time corresponding to a quarter note of the music) are increased. In addition to the above four types of feature values, the beat period ratio P5 Parameters such as the beat intensity P6 and the beat intensity ratio P7 may also be used as feature amounts. In extracting the feature amount from the target music piece, the feature amount extraction range may be extracted from an area composed of any combination of the whole music piece, a part of the music piece, and a plurality of parts of the music piece. In addition, as another method for obtaining the feature amount in step S10, the user's subjective impression on the music piece is obtained in advance as a plurality of feature quantities for the music piece by a sensory evaluation experiment using the SD (semantic differential) method or the like. A method may be used in which a database is created in association with the identifier, and the feature amount corresponding to the designated song is searched and obtained using the song identifier as a key. In step S11, the feature quantity obtained in the above-described step S10 is stored and managed in the feature quantity DB 21 by the feature quantity management means 11 in association with the song identifier of the corresponding song. An example of the content of the feature DB 21 is shown in FIG. The feature amount DB 21 may simultaneously manage information related to the music identifier, the singer name of the music, and the music location in the music information 32 format.
[0041]
On the other hand, in step S12, the reproduction detection means 14 detects that the user has reproduced the music, and notifies the reproduction history management means 15 of it. In step S13, the reproduction history management means 15 accesses the reproduction count DB 22 in accordance with the notification, and adds 1 to the reproduction count corresponding to the song identifier of the corresponding song. In this case, the amount to be added may be calculated by a separately determined coefficient. Further, in step S14, the preference level input means 12 causes the user to input a preference level F, which is the user's own preference level for a predetermined piece of music. In the present embodiment, it is assumed that the preference degree F is 0 or a positive value, but it is also possible to express like (positive) or dislike (negative) by making the value range positive or negative. Further, in the preference level input means 12, the user himself / herself inputs the preference level F, but the degree of preference for the music designated by the user is separately stored in advance and the data can be used. For this, automatic input may be performed with appropriate processing. In step S15, the input preference degree F is written in the preference level DB 23 by the preference level management means 13 in association with the music identifier. The reproduction count DB 22 and the preference level DB 23 may be managed as a single database using the format of the personal preference information 33 shown in FIG.
[0042]
When the recommended music presentation is requested by the user, the preference level and / or the number of reproductions of each music are extracted in step S16. This is because the recommendation degree calculation means 16 uses the data in the reproduction number DB 22 managed by the reproduction history management means 15 to the reproduction number Cn corresponding to the music identifier n and / or the preference degree DB 23 managed by the preference degree management means 15. This is done by extracting the preference degree Fn corresponding to the music identifier n from the data.
[0043]
In step S17, the recommended degree calculating means 16 calculates the primary recommended degree FCn for each piece of music registered in each DB.
[0044]
[Expression 1]

[0045]
(A1 and a2 are coefficients for adjusting the degree of preference and the degree of influence of the number of playbacks on the primary recommendation level FCn)
Further, only the number y of songs separately determined in descending order of the value of FCn is extracted by pairing each FCn and the corresponding song identifier n and output to the similar song search means 17. A set of elements obtained by pairing the music identifier n and the primary recommendation level FCn output at this time is defined as a primary recommendation level array {FCX}. The number of elements of the primary recommendation degree array {FCX} is y.
[0046]
In step S18, the similar music search means 17 searches for the music identifier of the similar music for each music identifier n that is an element of the primary recommendation degree array {FCX}.
[0047]
Next, a method for searching for similar songs will be described. Similar tunes are obtained in order for the elements of the primary recommendation degree array {FCX}. When the feature amount Pn (Pn1, Pn2, Pn3,... PnN) of a certain element (music identifier n) is given, the feature amount Pm (Pm1, Pm2, Pm2, Pm2, Pm2,... PnN) of the song identifier m of each song stored in the feature amount DB 21 Pm3... PmN) and the distance between the feature quantity Pn (Pn1, Pn2, Pn3,... PnN) of the music identifier n given as the search key are calculated as Euclidean distances between the feature quantities in Equation 2, The reciprocal is sequentially obtained as a value Rnm representing the degree of similarity.
[0048]
[Expression 2]

[0049]
(N is the number of feature quantities used in the present embodiment, Bi is a coefficient for normalizing each feature quantity, n is the nth element of the primary recommendation degree array {FCX}, and m is the feature quantity DB 21. (This corresponds to the m-th element of the music stored in.)
Here, for the calculation of Rnm, for n, the music of all elements of the primary recommendation degree array {FCX} is the calculation target, and for m, all the music stored in the feature value DB 21 is the calculation target. The value Rnm representing the similarity is calculated.
[0050]
Next, in step S19, the recommended music presentation means 18 calculates the secondary recommended degree RECnx by the following method.
[0051]
First, a similar music for the music identifier n which is the nth element of the primary recommendation degree array array {FCX} having the number of elements y is obtained. This is created as an array {TEMPn} in which the music identifier m, which is a similar music of the music identifier n, is associated with the similarity Rnm as an element of the array, and is arranged by the number z of songs separately determined in descending order of Rnm. . The number of elements in the array {TEMPn} is z. Here, the array {TEMPn} is created for each element of the primary recommendation degree array array {FCX}. In addition, the xth element of the array {TEMPn} is referred to as a music identifier nx and a similarity Rnx. The similarity Rnx between the music identifier n and the music identifier nx is extremely large if the Euclidean distance between the feature quantities is very close. Therefore, as an example, the similarity Rnx is converted into the number of approximations Snx. Determined by 3. The approximation degree Snx is a numerical value that represents the degree of approximation between music pieces that changes between 1 and 0.
[0052]
[Equation 3]

[0053]
(D1 and d2 are constants for adjusting the degree of approximation. In this example, d1 and d2 are positive values)
Since the primary recommendation degree of the music identifier n has already been obtained as FCn, the secondary recommendation degree RECnx of the music identifier nx is obtained by Equation 4 as follows.
[0054]
[Expression 4]

[0055]
Next, an array {TEMPRn} is defined as an element of the array {TEMPn} in which the similarity Rnx is replaced with the secondary recommendation RECnx among the pair of the music identifier nx and the similarity Rnx. The array {TEMPRn} has z elements.
[0056]
Thereafter, y elements {TEMPn} with z elements calculated from y elements of the array {FCX} are sequentially calculated, and the result is a secondary recommendation degree that is a two-dimensional (y, z) array. The array is {RECnx}. The secondary recommendation level array {RECnx} is a two-dimensional array composed of elements in which the secondary recommendation level RECnx and the music identifier j corresponding thereto are paired.
[0057]
In step S20, bibliographic information such as the song name is acquired based on the song identifier j of each element of the secondary recommendation degree array {RECnx} calculated by the recommended song presentation means 18, and the secondary recommendation degree RECnx is obtained. Present to the user in descending order.
[0058]
The above is an example of the calculation of the secondary recommendation degree RECnx and the secondary recommendation degree array {RECnx}, and the array {FCX} based on the calculation of the array {TEMPRn} at the head of the array {TEMPRn}. Is added as an element corresponding to FCn instead of the secondary recommendation RECnx as the value of the secondary recommendation corresponding to the music identifier n, and the number of elements of the array {TEMPRn} (Z + 1) may be used to determine the secondary recommendation degree array {RECnx}.
[0059]
Moreover, although the number of elements y and z of the array is a constant, it may be a variable number for each array when calculating each array. The primary recommendation degree FCn is calculated from the preference degree F and the number of reproductions C for each piece of music. However, separately, the user inputs a degree of preference for a music bibliographic item such as a favorite artist or genre, and It may be reflected in the degree of preference. In this case, if a preference level of a favorite genre is input, it is conceivable that the preference level for music belonging to the genre is automatically added by a coefficient e1. In addition, the recommended music presented to the user by the recommended music presentation means 18 may be extracted from the secondary recommendation level array {RECnx} of the above-described embodiment, or may be extracted from the primary recommendation level array {FCX}. Also good. In addition, among each piece of music that is an element of the primary recommendation degree array {FCX} calculated from the preference level and / or the number of times of reproduction, the music whose number of times of reproduction is a constant f1 or less that is determined separately is particularly preferentially extracted. good. As a result, the user can receive a preferential presentation from songs similar to the favorite song that have not yet been sampled or have not yet been sampled.
[0060]
In the present embodiment, it is assumed that the preference F is 0 or a positive value. However, by expressing the range of values as positive or negative, we like to express likes (positive) and dislikes (negative). Also good. This makes it possible to present recommended music for trial listening (as a change of mood) even for music that was previously classified as disliked.
[0061]
In addition, the user can change the degree of preference for the music according to the mood at that time, and the primary recommended degree FCn is not only the number of times the music is played, but also the number of times of the most recent period and the time change of the number of times the period is played. By adding it as a parameter at the time of calculation, the recommended music calculated by the secondary recommendation degree array {RECnx} or the primary recommendation degree array {FCX} can be changed according to the changing user's mood. It may be calculated. Furthermore, by making it possible to adjust the coefficients a1 and a2 of Formula 1 and the coefficients d1 and d2 of Formula 3 or the coefficients e1 and f1 in the above-described embodiment, calculation of recommended music that more suits the user's preference. May be possible.
[0062]
Further, the above embodiment may be implemented as a function completed by a single personal computer or an apparatus such as an audio player, or in an environment where they are connected via a network, each of the means 10 to 18 in FIG. 4 may be distributed or partially shared. Further, as shown in FIG. 4, a music service provider 41 on the Internet is provided with a feature music identifier and a music content database 42, a personal computer, etc. The user terminal 43 and the audio player 431 may be implemented with the music identifier preference playback count database 44, and each component may be implemented in another form, that is, on the Internet, by each means 10 of FIG. -18 may be implemented in any form distributed or partially shared.
[0063]
When the music service provider 41 sends / distributes music data to the user, the feature quantity and / or music information corresponding to the music data is associated with or separately from the music data of the music, and the components shown in FIG. The music service is sent to a certain user terminal 43, and the user terminal 43 and the audio player 431 return the user's preference for music and the number of reproductions to the music service provider 41 as needed. The provider 41 may be configured to be able to recommend sales songs that reflect user preferences.
[0064]
The music service provider 41 cooperates with an existing content provider to provide at least the function of the recommended music presentation means 18 to the user, so that the user can use the search fee and introduce the music to the content provider. It may be in the form of charging a fee / sales promotion fee, etc. and making it income. In addition, in the above-described embodiment, a function for executing at least the preference level calculation unit 12 and / or the reproduction detection unit 14 on the user terminal side such as a PC is provided to the user for a fee or free of charge. A license fee may be collected from the user according to the degree of use.
[0065]
【The invention's effect】
According to the present invention, when a user searches for a new piece of music that the user wants to obtain, or in a situation where a user owns a large number of songs and the user himself or herself cannot grasp the music he owns, This makes it possible to recommend and present recommended music with higher accuracy. According to the present invention, the user simply inputs the degree of preference for a certain number of songs according to the initial stage or usage as the preference level. The search function and the recommendation level calculation function make it possible to extract and present music that suits the user's preference with high accuracy. Also, the longer you use it, the personal preference level and the number of playbacks will be accumulated as personal preference information, which can be extracted and presented with higher accuracy, and changes in user preferences will be reflected in the personal preference information. It is possible to present recommended songs according to changes in preferences. It is also possible to preferentially present a song that is similar to the user's preference from among the enormous number of songs but has never been sampled by the user.
[0066]
In addition, the search service provider simply extracts the feature amount of the music owned by the existing content data provider and registers it together with other information such as the music ID, so that the music provided by the content data provider can be obtained. Can be used as a search target for users, and by using the user's personal preference information, it is possible to present recommended songs with various functions as described above according to user preferences and changes in preferences. Become.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a schematic configuration of a music search system according to an embodiment of the present invention.
FIG. 2 is a flowchart used for explaining the operation of the music search system according to the embodiment of the present invention.
FIG. 3 is a diagram showing an example of a record handled in the music search system according to the embodiment of the present invention.
FIG. 4 is a system diagram showing an outline of an example of a music search system according to an embodiment of the present invention.
[Explanation of symbols]
10 Feature extraction means
11 Feature quantity management means
12 Preference input means
13 Preference level management means
14 Reproduction detection means
15 Playback history management means
16 Recommendation level calculation means
17 Similar song search means
18 Recommended song presentation means
21 Feature DB
22 Playback DB
23 Preference level DB
31 Feature data
32 Music information
33 Personal preference information
41 Music service provider
42 feature amount music identifier and music content database
43 User terminal
431 Audio Player
44 Music identifier preference playback count database
45 Internet

Claims

A feature amount managing means for managing a feature amount representing a feature of a song in association with a song identifier for identifying the song, a preference degree specified by a user for the song, and a song identifier for identifying the song A preference level managing means for managing in association, a recommendation level calculating means for calculating a recommendation level 1 based on the preference level associated with the music identifier, and extracting a music identifier of a recommended song for the user; Similar music search means for extracting a music identifier of a similar music corresponding to the recommended music from the feature quantity management means based on the music identifier of the recommended music, the recommended degree of recommendation music 1 and / or the similarity music A music search system comprising a recommended music presentation means for calculating a new recommended degree of 2 from the similarity and calculating and presenting a recommended music to the user.

Corresponding feature quantity management means that associates and manages a feature quantity that expresses the feature of a song and a song identifier that identifies the song, and the number of times the user has played the song and the song identifier that identifies the song A playback history management means for managing the recommendation, a recommendation degree calculating means for calculating a recommendation degree 1 based on the number of times of reproduction associated with the song identifier, and extracting a song identifier of a recommended song for the user, and the recommendation Similar song search means for extracting a song identifier of a similar song with respect to the recommended song from the feature quantity management means based on the song identifier of the song, and having a recommendation degree of the recommended song of 1 and / or similarity of the similar song A music search system characterized by comprising a recommended song presentation means for calculating a new recommendation degree 2 from the degree and calculating and presenting a recommended song for the user.

A feature amount managing means for managing a feature amount representing a feature of a song in association with a song identifier for identifying the song, a preference degree specified by a user for the song, and a song identifier for identifying the song Preference level managing means for managing in association with each other, reproduction history managing means for managing in association with the number of times the user has played the music and the music identifier for identifying the music, and the music identifier A recommendation degree calculating means for calculating a recommendation degree 1 based on the preference degree and / or the number of times of reproduction, and extracting a song identifier of a recommended song for a user, and the feature quantity managing means based on the song identifier of the recommended song Similar song search means for extracting a song identifier of a similar song to the recommended song from the recommended song, and calculating a new recommendation degree 2 from the recommendation degree 1 of the recommended song and / or similarity of the similar song, Guess against Music search system characterized in that it consists of recommended songs presenting means for calculating presenting songs.

A preference level management means for managing a preference level specified by a user for a song and a song identifier for identifying the song, a number of times the user has played the song, and a song identifier for identifying the song And a history of reproduction management means for managing and a recommendation degree 1 based on the preference degree and the number of times of reproduction associated with the music identifier, and a recommendation degree for extracting a music identifier of a recommended song for the user A music search system comprising: a recommended music presentation means for calculating and presenting a recommended music to the user based on a recommendation degree 1 of the recommended music.

The preference level management means uses, as a preference level, a value specified by a user for a song and / or a user's preference level for music bibliographic information such as a favorite artist or genre by a predetermined coefficient. The music search system according to any one of claims 1 to 4, wherein the music search system according to any one of claims 1 to 4 is integrated.

When the recommended song presenting means calculates and presents a recommended song, the number of plays of the recommended song is not more than a numerical value separately determined and / or a preference is not more than a numerical value separately determined, and is preferentially presented. The music search system according to any one of claims 1 to 5.

The preference level management unit automatically acquires a degree of preference of a user's music from a database created in advance or separately, and manages the degree of preference in association with a music identifier for identifying the music. The music search system according to any one of 1 to 6.

The reproduction history management means manages the date and time when the user reproduces the music and / or the history of the number of reproductions within a certain period and / or the history of the number of reproductions within a certain period and the music identifier for identifying the music in association with each other. The music search system according to any one of claims 2 to 7, wherein:

Arbitrary means of the feature quantity management means, the preference degree management means, the reproduction history management means, the recommendation degree calculation means, the similar song search means, and the recommended song presentation means are between the related other means. 9. The music search system according to claim 1, wherein input / output of processing information is processed via a network and / or the Internet.

A music search method for searching for a recommended song for a user, a feature quantity management database that manages a feature quantity that expresses a feature of a song and a song identifier that identifies the song, and a user for the song A preference degree management database that manages a preference degree to be specified and a music identifier for identifying the music in association with each other, calculates a music recommendation degree 1 from the preference degree management database, and uses the recommendation degree calculation result Extract similar songs and their similarities of the music from the feature quantity management database, calculate a new recommendation level 2 using the similarity extraction result and / or the recommendation level calculation result, and calculate a recommended song for the user A music search method characterized by presenting.

A music search method for searching for a recommended song for a user, a feature quantity management database that manages a feature quantity that expresses a feature of a song and a song identifier that identifies the song, and a user for the song A reproduction history management database that manages the number of reproduced reproductions and a music identifier for identifying the music in association with each other, calculates a recommendation degree 1 of the music from the reproduction history management database, and uses the recommendation degree calculation result Extract similar songs and their similarities of the music from the feature quantity management database, calculate a new recommendation level 2 using the similarity extraction result and / or the recommendation level calculation result, and calculate a recommended song for the user A music search method characterized by presenting.

A music search method for searching for a recommended song for a user, a feature quantity management database that manages a feature quantity that expresses a feature of a song and a song identifier that identifies the song, and a user for the song A preference management database that manages a preference level to be specified and a song identifier that identifies the song, and a playback that manages the song number that identifies the song and the number of times the user has played the song. Having a history management database, calculating a recommendation level 1 of the music from the preference level management database and / or the playback history management database, and using the recommendation level calculation result, the similar music of the music from the feature amount management database The similarity is extracted, a new recommendation level 2 is calculated using the similarity extraction result and / or the recommendation level calculation result, and Music search method and calculates presenting Susumukyoku.

A music search method for searching for a recommended song for a user, a preference degree management database that manages a preference degree specified by a user for a song and a song identifier that identifies the song, and a song A reproduction history management database that manages the number of reproductions played by the user in association with a music identifier that identifies the music, and calculates a recommended degree of music 1 from the preference management database and the reproduction history management database A music search method comprising calculating and presenting a recommended song for a user based on a recommendation degree 1 of the recommended song.

When calculating the degree of preference, the value specified by the user for the music and / or the degree of the user's preference for the bibliographic information of the music such as the favorite artist and genre are integrated by a predetermined coefficient when calculating the degree of preference. The music search method according to any one of claims 10 to 13.

15. When a recommended song is calculated and presented, a song whose number of reproductions of the recommended song is equal to or less than a separately determined value and / or a preference value or less is separately specified is preferentially presented. The music search method according to any one of the above.

11. The degree of preference of a user's music is automatically acquired from a database created in advance or separately when calculating the degree of preference, and managed in association with a music identifier for identifying the music. The music search method according to any one of 15.

The reproduction history management database associates and manages the date and time when the user reproduces the music and / or the history of the number of times of reproduction within a certain period and / or the history of the number of times of reproduction within a certain period, and the music identifier for identifying the music. The music searching method according to any one of claims 11 to 16, wherein the music searching method is any one of the above.

18. The feature quantity management database, the preference degree management database, and the reproduction history management database are processed individually or in an arbitrary combination on the network and / or the Internet. The music search method according to claim 1.

A feature amount managing means for managing a feature amount representing a feature of a song in association with a song identifier for identifying the song, a preference degree specified by a user for the song, and a song identifier for identifying the song Preference level management means for managing in association;
Based on the preference degree associated with the music identifier, the recommendation degree 1 is calculated, the recommendation degree calculating means for extracting the music identifier of the recommended music for the user, and the music identifier of the recommended music, Similar music search means for extracting music identifiers of similar music to the recommended music from the feature quantity management means, and calculating a new recommendation level 2 from the recommendation level 1 of the recommended music and / or the similarity level of the similar music And a recommended music presenting means for calculating and presenting recommended music for the user.

Corresponding feature quantity management means for managing the feature quantity expressing the feature of the song and the song identifier for identifying the song in association with each other, the number of times the user has played the song and the song identifier for identifying the song Replay history management means to manage,
A recommendation degree 1 is calculated based on the number of times of reproduction associated with the song identifier, and a recommendation degree calculating means for extracting a song identifier of a recommended song for a user, and based on the song identifier of the recommended song, Similar music search means for extracting a music identifier of a similar music for the recommended music from the feature quantity management means, and calculating a new recommendation level 2 from the recommendation level 1 of the recommended music and / or the similarity level of the similar music And a recommended music presenting means for calculating and presenting recommended music for the user.

A feature amount managing means for managing a feature amount representing a feature of a song in association with a song identifier for identifying the song, a preference degree specified by a user for the song, and a song identifier for identifying the song Preference level management means for managing in association;
Recommended based on the reproduction history management means for managing the number of times the user has played the music in association with the music identifier for identifying the music, and the preference level and / or the number of times of playback associated with the music identifier 1 is calculated, and a recommendation degree calculating means for extracting a song identifier of a recommended song for the user, and a music identifier of a similar song to the recommended song from the feature amount management means based on the song identifier of the recommended song. A recommended song presenting means for calculating a recommended song for the user by calculating a new recommended degree from the recommendation degree of the recommended song and / or the similarity of the similar song. A music search apparatus characterized by comprising:

A preference level management means for managing a preference level specified by a user for a song and a song identifier for identifying the song, a number of times the user has played the song, and a song identifier for identifying the song And a history of reproduction management means for managing and a recommendation degree 1 based on the preference degree and the number of times of reproduction associated with the music identifier, and a recommendation degree for extracting a music identifier of a recommended song for the user A music search apparatus comprising: a recommended music presentation means for calculating and presenting a recommended music for a user from a recommendation degree 1 of the recommended music.

The preference level management means uses, as a preference level, a value specified by a user for a song and / or a user's preference level for music bibliographic information such as a favorite artist or genre by a predetermined coefficient. The music search device according to any one of claims 19 to 22, wherein

When the recommended song presenting means calculates and presents a recommended song, the number of plays of the recommended song is not more than a numerical value separately determined and / or a preference is not more than a numerical value separately determined, and is preferentially presented. The music search device according to any one of claims 19 to 23.

The preference level management unit automatically acquires a degree of preference of a user's music from a database created in advance or separately, and manages the degree of preference in association with a music identifier for identifying the music. The music search device according to any one of 19 to 24.

The reproduction history management means manages the date and time when the user reproduces the music and / or the history of the number of reproductions within a certain period and / or the history of the number of reproductions within a certain period and the music identifier for identifying the music in association with each other. The music search device according to any one of claims 20 to 25, wherein:

Arbitrary means of the feature quantity management means, the preference degree management means, the reproduction history management means, the recommendation degree calculation means, the similar song search means, and the recommended song presentation means are between the related other means. 27. The music search apparatus according to claim 19, wherein input / output of processing information is processed via a network and / or the Internet.

The operation of the music search system according to any one of claims 1 to 9, the function realized by the music search method according to any one of claims 10 to 18, or claims 19 to 27 A program for causing a computer to execute the operation of the music search device according to any one of the above.

A computer-readable recording medium on which the program according to claim 28 is recorded.