|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Weekly Report for Saturday February 21, 2004 Home Reports | ||||||||||||||||||||||||||||||||
| I. Meeting with Russ | ||||||||||||||||||||||||||||||||
| After talking with Russ it seemed apparent to me that he wanted to work on antenna/wireless communications, possibly without me and/or Will, He said he felt this way because he had a great degree of familiarity with the topic due to one of his classes, and constraints due to two out of the three members of our group commuting to school. | ||||||||||||||||||||||||||||||||
| II. Meeting with Dr. Tahernezhadi | ||||||||||||||||||||||||||||||||
| I explained to Dr. Tahernezhadi how Russ felt about the group project, and that he thought that I should not be working with him. I asked for permission to work alone and for a topic that I could do, was told that speech enhancement would be a good topic for me to do. | ||||||||||||||||||||||||||||||||
| I was given a general procedure for going about the project. Start by going taking a recorded voice signal and adding to it noise through use of the command "randn" in Matlab. Taking a Fast Fourier Transform (FFT) and then looking at the power of the specific frequency bins to determine if the bin contains noise or if it contains the desired signal. Then the attenuation of the ones that contain noise has to be done. Since only five frequency components of the voice exist (although varying) and noise spans the intire portion of the spectrum the majority of the bins will be attenuated. Naturally after this one must take an inverse Fourier Transform and then examine the modified signal. The only way to guess at what the power difference will be between the noise and the desired signal is to graph the power of the noise and compare it with a graph of the power of the signal to look at the difference in power. | ||||||||||||||||||||||||||||||||
| III. Dr. Quo's Speech | ||||||||||||||||||||||||||||||||
| I attended Dr. Quo's IEEE speech on active noise control. He discussed an array of topics there concerning noise control, including topics from our class such as echo cancellation and speech enhancement. Also afterword I discussed briefly with him the use of directional audio speakers and how they workd, a topic of much interest to me before. The way that is works is as follows, first you transfrom the signal to two higher frequencies. Then you broadcast and four frequencies will exist, the sum, the difference, and the two orignial frequencies. Only the difference can be heard by the human ear, or so that is desired effect. This works well in accoustic cancellation of noise in applications such as speaker phones where the original persons speach may be picked up in a full duplex system by the microphone at the other end, since it has a very narrow area that the sound is projected through, it will not go back to the microphone and will thus not cause interference with the people talking. | ||||||||||||||||||||||||||||||||
| IV. Research | ||||||||||||||||||||||||||||||||
| The following are links to some of my more valuable readings... | ||||||||||||||||||||||||||||||||
| DSP Voice Frequency Compander | ||||||||||||||||||||||||||||||||
| Summary: using filters and frequency shifting one can derive a method of reducing noise. | ||||||||||||||||||||||||||||||||
| Definition: voice frequency | ||||||||||||||||||||||||||||||||
| Summary: basically described the range of frequencies in which a voice exists (300Hz-3400Hz) and the general channel used in telephones (4kHz) for transmission. | ||||||||||||||||||||||||||||||||
| FFTW | ||||||||||||||||||||||||||||||||
| Summary: this is the web site for the "Fastest Fourier Transform in the West" and basically it has the coding in C for efficient fourier transforms which may prove useful in the future when it comes time to script the code for doing an FFT in this the procedure of noice reduction. | ||||||||||||||||||||||||||||||||
| Previous Report Page 2 of this Report Page 3 of this Report | ||||||||||||||||||||||||||||||||