GLOS: GLOve for Speech Recognition

摘要

Technological advancements in the field of Biomedical Engineering have allowed impaired individuals to use assistive devices in order to improve their quality of life. In the case of deafblind subjects, who experience both visual and auditory loss, the majority of available assistive devices are invasive (e.g. cochlear implants). Non-invasive technological improvements are extremely limited, in part due to the lack of scientific research interest in dual sensory loss [1]. In this paper we aim to present GLOS, a low-cost and non-invasive device that will allow the deafblind individuals to comprehend speeches in real-time. The proposed wearable device uses off-the-shelf components such as a Raspberry Pi 3 board, a simple microphone, and haptic feedback vibrating disks. The recorded speech from the microphone is processed by the board and encoded into 5 haptic vibrating modules attached to a glove. Each haptic vibrating module is placed on a different finger of the right hand. The current available non-invasive solutions do not use speech and therefore they do not allow for live communication (e.g. MyVox [2], Sparsha [3] and Mobile Lorm Glove [4]) or require long procedures to convert the messages (e.g. Finger Braille Teaching System [5]). This new biomedical device aims at overcoming these limitations. The authors tested the device in a preliminary testing and it was shown that GLOS has an average accuracy of 91.67% when tested for the recognition of twenty encoded sentences. The authors were deprived of both visual and hearing inputs and were trained for half hour per day for a period of 30 days.