Research press release


Nature Communications

Engineering: Vocal vibrations used for voice recognition system



今回、Kilwon Choたちの研究グループは、正確な音声検出ができる皮膚貼り付け型フレキシブルセンサーを作製する方法を考案した。この設計により、減衰(振動の振幅の減少)が最小限に抑えられ、声の圧力(声のパターンと声量)と直接相関する測定値である皮膚の加速度を正確に検出できるようになった。Choたちは、今回設計された皮膚にフィットする極薄型電子デバイスが、人間の声の周波数帯域全体にわたって高感度で、正確に音声を検出できることを明らかにした。


A sensor that attaches to the skin that can accurately recognize the human voice by detecting vibrations on the skin of the neck is demonstrated in Nature Communications this week. As demonstrated in a series of proof-of-principle tests, the device has potential applications in voice-recognition security systems and in monitoring vocal health.

Skin-attachable vibration sensors that detect human voices have the potential to compete with existing speech recognition technologies. Despite advantages, such as being able to detect voices in noisy or windy environments, current skin-attachable sensors have performance limitations that hinder their use. These include inconsistencies in their sensitivity over the entire range of vocal frequencies.

Kilwon Cho and colleagues devised an approach for fabricating flexible, skin-attachable sensors capable of accurate voice detection. Their design minimized damping (a reduction in the amplitude of the vibrations) and allowed for accurate detection of skin acceleration - a measurement that directly correlates with voice pressure (vocal patterns and volume). The authors found that their ultrathin skin-conformable electronic device was able to accurately detect voices with high sensitivity over the entire frequency range of the human voice.

In tests using their device as part of a voice-recognition security system for entry to a room, the authors found that their system was capable of recognizing the correct user’s voice even when another user uttered the password. The system was also capable of recognizing the correct user’s voice clearly when they wore a face mask.

doi: 10.1038/s41467-019-10465-w

「Nature 関連誌注目のハイライト」は、ネイチャー広報部門が報道関係者向けに作成したリリースを翻訳したものです。より正確かつ詳細な情報が必要な場合には、必ず原著論文をご覧ください。

メールマガジンリストの「Nature 関連誌今週のハイライト」にチェックをいれていただきますと、毎週最新のNature 関連誌のハイライトを皆様にお届けいたします。