Real-Time Detection of Important Sounds with a Wearable Vibration Based Device for Hearing-Impaired People


YAĞANOĞLU M., KÖSE C.

ELECTRONICS, cilt.7, sa.4, 2018 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 7 Sayı: 4
  • Basım Tarihi: 2018
  • Doi Numarası: 10.3390/electronics7040050
  • Dergi Adı: ELECTRONICS
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Anahtar Kelimeler: wearable processing, real-time systems, quality of life, feature extraction, speech processing, audio fingerprint, AUDIO CLASSIFICATION, NEURAL-NETWORKS, RECOGNITION, AUGMENTATION, SYSTEMS
  • Atatürk Üniversitesi Adresli: Evet

Özet

Hearing-impaired people do not hear indoor and outdoor environment sounds, which are important for them both at home and outside. By means of a wearable device that we have developed, a hearing-impaired person will be informed of important sounds through vibrations, thereby understanding what kind of sound it is. Our system, which operates in real time, can achieve a success rate of 98% when estimating a door bell ringing sound, 99% success identifying an alarm sound, 99% success identifying a phone ringing, 91% success identifying honking, 93% success identifying brake sounds, 96% success identifying dog sounds, 97% success identifying human voice, and 96% success identifying other sounds using the audio fingerprint method. Audio fingerprint is a brief summary of an audio file, perceptively summarizing a piece of audio content. In this study, our wearable device is tested 100 times a day for 100 days on five deaf persons and 50 persons with normal hearing whose ears were covered by earphones that provided wind sounds. This study aims to improve the quality of life of deaf persons, and provide them a more prosperous life. In the questionnaire performed, deaf people rate the clarity of the system at 90%, usefulness at 97%, and the likelihood of using this device again at 100%.