INFORMATION PREDICTION SYSTEM, INFORMATION PREDICTION METHOD AND INFORMATION PREDICTION PROGRAM

• Main Authors: SANO AKIRA, HAGIWARA TORU, HISAKU AKIYOSHI, NAGATA TAKESHI, MATSUZAKI KAZUTOSHI, MAEKAWA HIDEMASA, TOMOSAWA HIROMITSU, MIZUTANI MAKIKO
• Format: Patent
• Language: English; Japanese
• Published: 26.05.2016
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PICTORIAL COMMUNICATION, e.g. TELEVISION

PROBLEM TO BE SOLVED: To provide an information prediction system, information prediction method and information prediction program for efficiently predicting a second feature vector on the basis of a first feature vector.SOLUTION: An image processing system 20 includes a face image storage unit 22 for recording a learning object, a learning result storage unit 23 for recording a learning result, and a control unit 21 connected to an input part. Then, the control unit 21 generates a connection main component base vector and an average vector by generating a row vector obtained by dividing main component coefficient vectors of a first feature vector and a second feature vector calculated in each learning object by an index representing variation of the main component coefficient vector and performing a main component analysis, and records the connection main component base vector and the average vector in the learning result storage part 23. Then, the main component coefficient vector of the first feature vector of a processing object, and the connection main component base vector recorded in the learning result storage part 23 are used, a main component coefficient vector of the second feature vector is calculated, and a second feature vector of the processing object is calculated.SELECTED DRAWING: Figure 1 【課題】第１の特徴ベクトルに基づいて、第２の特徴ベクトルを効率的に予測するための情報予測システム、情報予測方法及び情報予測プログラムを提供する。【解決手段】画像処理システム２０は、学習対象を記録する顔画像記憶部２２と、学習結果を記録する学習結果記憶部２３と、入力部に接続された制御部２１とを備える。そして、制御部２１は、学習対象毎に、算出した第１の特徴ベクトル及び第２の特徴ベクトルの主成分係数ベクトルを、主成分係数ベクトルのばらつきを表わす指標で除算した行ベクトルを生成し、主成分分析を行なうことにより、結合主成分基底ベクトル、平均ベクトルを生成し、学習結果記憶部２３に記録する。そして、処理対象の第１の特徴ベクトルの主成分係数ベクトル、学習結果記憶部２３に記録された結合主成分基底ベクトルを用いて、第２の特徴ベクトルの主成分係数ベクトルを算出し、処理対象の第２の特徴ベクトルを算出する。【選択図】図１