Speech recognition in loud noise using an ear-insert microphone

Think A Move, Ltd. has developed a patented ear-insert microphone which captures speech as acoustic vibrations inside the ear canal. These vibrations propagate to the ear canal through the flesh and bones in the human skull. A high density foam on the earpiece seals the ear canal when the earpiece i...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:The Journal of the Acoustical Society of America 2010-03, Vol.127 (3_Supplement), p.1815-1815
Hauptverfasser: Pruthi, Tarun, Despa, Mihai, Juneja, Amit
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Think A Move, Ltd. has developed a patented ear-insert microphone which captures speech as acoustic vibrations inside the ear canal. These vibrations propagate to the ear canal through the flesh and bones in the human skull. A high density foam on the earpiece seals the ear canal when the earpiece is inserted. Tests show that this earpiece provides an average passive noise cancellation (PNC) of around 38 dB for noises in Aurora database. Using an in-house speech command recognizer, with a short enrollment phase, on a database of 19 speakers (11 females, 8 males), speaking a vocabulary of 56 commands, an average accuracy of 85% has been observed in 90 dBA of tank, military vehicle, and machine gun noises. To further demonstrate the noise robustness of the earpiece as compared to external microphones, pilot tests were conducted on a small set of speakers to recognize speech commands recorded simultaneously with an external microphone and our ear-insert microphone with our recognizer. Results show that while the accuracy of the recognizer drops to 27% in 90 dBA noise from 96% in quiet for external microphone, it only drops to 92% in 90 dBA of noise from 95% in quiet for internal microphone.
ISSN:0001-4966
1520-8524
DOI:10.1121/1.3384184