Abstract
Hearing impairment can significantly reduce the quality of life of an individual or even create hazardous situations. The development of aiding devices for automated sound recognition is becoming increasingly important. This research proposes a sound recognition device dedicated to environmental sounds, designed with main focus on portability and user-friendliness. The pre-processing stage is loosely based on the human hearing system, while a robust set of binary features permits a simple classification strategy to be used. This gives the system the capability of in-field learning, greatly improving its usability. The system was implemented in an FPGA based platform, developed in-house specifically for this application. The device does not require any critical parameter calibration nor any technical expertise from the user. The prototype also presents an approachable user interface, a key aspect for elderly users. Sensitivity and specificity were evaluated over several random subsets of signals. The system presented high accuracy, and even though a few sounds were difficult to recognize at higher levels of noise, there was a very small number of false-positives.
