Biography
Enrollment Date: 2010
Graduation Date：2013
Degree：M.S.
Defense Date：2012.12.17
Advisors：Dongmei Li
Department：Institute of Microelectronics，Tsinghua University
Title of Dissertation/Thesis：Environment Noise Cancellation Algorithm Research Of An Electronic Stethoscope With Two Microphones
Abstract:
More and more people have cardiopulmonary diseases all over the world, stethoscope has an irreplaceable place in diagnosis these diseases in an early stage. With the development of technology, electronic stethoscope with the function of recording, amplifying, and filtering, is needed for telemedicine. However, it amplifies environment noise as it amplifies needed sound, which makes doctors uncomfortable, and prevent it from widely use.
In this paper, we discussed some common methods of noise cancellation, such as, low pass filter, wavelet, adaptive filtering, etc. And promoted a novel method based on CASA (Computational Auditory Scene Analysis). We compared it to NLMS in detail, as they both use tow-microphone. The measurement standard is Signal-to-Noise Ratio.
CASA uses the energy difference between the signals from the two microphones, the inner and the outer. The signal is divided by frame and by sub-band, into items, the feature-energy difference-of each item is calculated, called IID. If it is above the threshold, we keep the item because signal is more than noise. If not, we cancel it by multiplying a very small number. At last, the weighed signal is send out.
NLMS is the mostly used adaptive filtering algorithm, called Normalized Least Mean Square. It simulates the travelling path of the noise, minus the estimated noise with real noise, and calculates its least mean square. Then adjust the coefficients of the filter, making it the best one in a sense.
At last, we tried to use OLA by FFT, to replace the GAMMATONE sub-band function of CASA, so that we can make it possible to run on a general DSP.