This might be the last report I wrote for this project although I would like to keep working on this during the coming weeks and months. Good news was that we make big progress during the past period.
Since we were stuck somehow, Mr. Richter Ed was intended to search for help from other people. So he asked me to modify the Matlab code as to make it easy to read, understand and test. After week 7, the code was significantly improved. The whole structure was slightly changed for the easiness to read and understand. New scripts and functions were also applied to make the program much smarter and more logical. A script, configuration.m where most of constant and global variables were assigned, was introduced for the consistency and convenience during test.
In week 8, a new member Ricky joined my work. He was a senior from Tsinghua University. I am really glad to work with a talent teammate from the top university in China. He suggested using cross correlation in time domain to compute the lag time for each of 64 microphones. His idea did help. This new lag timetable on which shifting is based is dynamic and with high accuracy. The corresponding result is much faster and better. (Pic. 1)
Thanks for the improvement we had in week 8, we started over the SNR test in week 9 and found that the terrible performance of SNR might be partly due to the omission of the power of harmonic components of the signal. The main factor of the unexpected SNR, we supposed, was that there were noises at certain frequencies in the background noise. Those noises at certain frequencies were opposite to white Gaussian noise. Their power would be added constructively during beamforming, and leading to higher power gain than expected. Ricky and I had raised the SNR by filtering those noises at certain frequencies (Pic. 2). But there were still several questions we were incapable to answer: 1. Where do the noises come from? 2. Why does the frequency of those noises shift?
I was supposed to work on this project for 8 weeks while this was the end of the 9th week. My course process might be over next week, but I will keep working on this project during the rest of my summer vocation, even after new semester begins.
Since we were stuck somehow, Mr. Richter Ed was intended to search for help from other people. So he asked me to modify the Matlab code as to make it easy to read, understand and test. After week 7, the code was significantly improved. The whole structure was slightly changed for the easiness to read and understand. New scripts and functions were also applied to make the program much smarter and more logical. A script, configuration.m where most of constant and global variables were assigned, was introduced for the consistency and convenience during test.
In week 8, a new member Ricky joined my work. He was a senior from Tsinghua University. I am really glad to work with a talent teammate from the top university in China. He suggested using cross correlation in time domain to compute the lag time for each of 64 microphones. His idea did help. This new lag timetable on which shifting is based is dynamic and with high accuracy. The corresponding result is much faster and better. (Pic. 1)
Thanks for the improvement we had in week 8, we started over the SNR test in week 9 and found that the terrible performance of SNR might be partly due to the omission of the power of harmonic components of the signal. The main factor of the unexpected SNR, we supposed, was that there were noises at certain frequencies in the background noise. Those noises at certain frequencies were opposite to white Gaussian noise. Their power would be added constructively during beamforming, and leading to higher power gain than expected. Ricky and I had raised the SNR by filtering those noises at certain frequencies (Pic. 2). But there were still several questions we were incapable to answer: 1. Where do the noises come from? 2. Why does the frequency of those noises shift?
I was supposed to work on this project for 8 weeks while this was the end of the 9th week. My course process might be over next week, but I will keep working on this project during the rest of my summer vocation, even after new semester begins.