THESIS
2016
xiii, 89 pages : illustrations ; 30 cm
Abstract
Amplifying both the auditory noise and signals will not help a hearing impaired person to hear the signals. This thesis designs, develops and prototypes an improved method of separating a mixture of audio signals into target signal(s) of interest and other undesired one(s) for hearing devices or systems. This method enables hearing devices to suppress the noise and selectively amplify the signals. Our solution uses 4
th order statistics to separate the audio signals. In particular, our novel
algorithms tackle and solve the asynchronization problems that make traditional source separation
algorithms useless. Traditional source separation algorithms use at least two microphones and asynchronization of signals is inevitable due to different propagation times to different microphones. In t...[
Read more ]
Amplifying both the auditory noise and signals will not help a hearing impaired person to hear the signals. This thesis designs, develops and prototypes an improved method of separating a mixture of audio signals into target signal(s) of interest and other undesired one(s) for hearing devices or systems. This method enables hearing devices to suppress the noise and selectively amplify the signals. Our solution uses 4
th order statistics to separate the audio signals. In particular, our novel
algorithms tackle and solve the asynchronization problems that make traditional source separation
algorithms useless. Traditional source separation algorithms use at least two microphones and asynchronization of signals is inevitable due to different propagation times to different microphones. In this thesis, relative propagation delays are estimated in real-time using pattern recognition algorithms and a novel synchronization processing algorithm. The estimated delays are then applied to resynchronize the signals before feeding them to the source separation algorithms. A successful simulation and a workable prototype are illustrated in this thesis. This original and novel technology has been selected to be patented by HKUST. In addition, a business plan has been drawn up and a startup company called ‘Incus Company Ltd.’ has been registered in Hong Kong. The earlier business plan won the 1
st runner-up award in 2015’s HKUST MBA Business Plan Competition and entered the final eight in the HKUST One Million Dollars Competition.
Post a Comment