TRANSMITTER COLLATION DEVICE, LEARNING DEVICE, TRANSMITTER COLLATION METHOD, LEARNING METHOD, AND PROGRAM

• Main Authors: OTSUJI TAICHI, TAKEUCHI TOSHIKI, AKIBA HIROYUKI
• Format: Patent
• Language: English; Japanese
• Published: 21.06.2022
• Subjects: ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; TRANSMISSION

To reduce situations in which in the case of estimating, from a signal wirelessly transmitted from a transmitter of a collation object, a plurality of types of information indicating the transmitter, mismatch between estimated results is caused.SOLUTION: A transmitter collation device 1 comprises a receiving unit 1a, radio wave feature amount generation unit 1b, a storage unit 1c, a similarity calculation unit 1d, an addition similarity calculation unit 1e, and an inference unit 1f. The storage unit 1c stores K (integer equal to or greater than 2) sets of template feature amount groups for inferring K types of information showing a transmitter. The similarity calculation unit 1d generates an i(integer of 1 to K)-th sample feature amount from a radio wave feature amount and calculates an i-th similarity group on the basis of the i-th sample feature amount and i-th set of template feature amount group. The addition similarity calculation unit 1e adds K similarities corresponding to i being 1 to K using an i-th weighting coefficient to calculate an addition similarity. The inference unit 1f infers K pieces of information previously associated with respective calculation sources of K similarities maximizing the addition similarity to be information indicating the transmitter.SELECTED DRAWING: Figure 1 【課題】照合対象の送信装置から無線送信される信号からその装置を示す複数種類の情報を推定する場合に推定結果が不整合となる状況を低減させる。【解決手段】送信装置照合装置１は、受信部１ａ、電波特徴量生成部１ｂ、記憶部１ｃ、類似度算出部１ｄ、加算類似度算出部１ｅ、推定部１ｆを備える。記憶部１ｃは送信装置を示すＫ（２以上の整数）種類の情報を推定するためのＫセットのテンプレート特徴量群を記憶する。類似度算出部１ｄは電波特徴量からｉ（１～Ｋの整数）番目のサンプル特徴量を生成しｉ番目のサンプル特徴量とｉセット目のテンプレート特徴量群とに基づきｉ番目の類似度群を算出する。加算類似度算出部１ｅはｉが１～ＫまでについてのＫ個の類似度をｉ番目の重み係数を用いて加算して加算類似度を算出する。推定部１ｆは加算類似度が最大となるＫ個の類似度のそれぞれの算出元に予め対応付けられたＫ個の情報を送信装置を示す情報と推定する。【選択図】図１