安全見守りシステムにおけるプライバシー保護手法

摘要

我々はこれまで，自己の経験や状況を伝えることが難しい就学前児童や高齢者にウェアラブルデバイスを装着し，デバイスに録音された装着者の声を家族が確認できる手法を構築してきた.本稿では，装着者の周囲にいた人物の会話が入力されることにより生じるプライバシーの問題から，録音データ中，装着者が発声した部分のみを家族に提供する.従来，発声を認識する手法では，特定の環境下における録音が必要であった.本手法では，周囲の音を抑制する差分マイクをデバイスに搭載し，装着者の発声が相対的に強調された録音データを獲得することで，録音環境の制約を設けることなく装着者の発声を検出できる.また，マイコンや無線LANモジュールを搭載したウェアラブルデバイスの実装により，保育園等の施設における装着者の発声を検出し，各家庭に送信するシステムを構築する.実装したデバイスによる録音データを用いてプライバシー保護の性能を評価した結果，装着者の周囲の音が有効に抑制されることを確認した.%Abstract We have developed a system of monitoring safety for preschoolers and senior citizens who cannot tell accurately about their own experiences and condition. A wearable device which has the recording capability attached to preschoolers and senior citizens, and their family can check the device wearer's voice recorded by the device. When the wearer's voice is recorded, conversation around the wearer might be input unintentionally. The privacy of the person who was inadvertently recorded may be violated. Therefore, this system provides the family with only the wearer's voices in the recorded data. Previously, there have been studies about recognizing voices, we must record in a particular condition. In this system, it is possible to detect the wearer's voices without constrained conditions because the wearable device has a differential microphone that can tone down ambient sounds, and the microphone can emphasize the wearer's voice more than the ambient sounds. Furthermore, it is possible to detect the wearer's voices in facilities (e.g. nursery school) and send detected voices to their homes by constructing the wearable device that has a microcomputer and a wireless LAN module. In the evaluation experiment, this system can effectively tone down ambient sounds.