Acoustic Vector Sensors Notes SP2010

Team Members: Evan

PhD Supervisor: Satya, Murat

Faculty Supervisor: Arye Nehorai

Goals

Understanding the AVS response:

1. To compute the frequency response and identify the bandwidth with flat response in all the channels.
2. To study the effects of the equalizer on the velocity measurements.

Perform 3D real-time source location using the four AVS

1. Revise the GUI
2. Implement an efficient method for triangulation in 3D

Status

Week 1 (1/25 and 1/27)

This week we worked out minor bugs in code and got everything back to where we left off last semester. The echo problem is still present. We discussed different ways to fix this problem. The assumptions that we made for the Capon spectrum to work may not be accurate (far field vs. near field acoustic waves)

Week 2 (2/1 and 2/3)

We gathered data from an AVS and compared it to the data we generate from the simulated data. The simulated data has the different components of the sensor in phase while the actual collected data did not. We altered the simulation to make the waves out of phase and observed the same 'echo' effect that we experience in the real data. This may be the source of the echo in the real data. Next week we will try to find a way to get the real data in phase.

Week 3 (2/8 and 2/10)

This week we developed a MATLAB script to correct the phase difference in the data. On Wednesday we finished the script and tested it on our phase shifted simulation. The script worked and lessened the echo to roughly 1/4 the magnitude of the real peak. Some LabView bugs made it difficult to continue working with the program, but hopefully they will resolve themselves when we switch computers. Next week we will attempt to apply the script to the real data- hopefully it will remove the echo that we previously experienced.

Week 4 (2/15 and 2/17)

We are still having a lot of trouble working out the problems with LabView. Sometimes matlab scripts which are changed while LabView is open do not cooperate and update well. We are trying to work past this. We are expanding the matlab script which removes the phase difference to shift multiple sets of data- hopefully in the near future this will be complete and we can test out our script on real data.

Week 5 (2/22 and 2/24)

Today we had promising progress with LabView- we now have the simulation for two sensors working again, although phase shifts to the data sometimes does not effect the right component of our estimated direction of arrival. On Wednesday we worked more with calibration and integrating the script into the real data acquisition.

Week 6 (3/1 and 3/3)

This week we finished integrating our cross correlation script that removed the phase shift from the data. The script seemed to be doing its job - a graph of the waves seemed to show that the script corrected any phase shifts. The phase correction did not remove the echo effect like we expected. What did work for the simulation did not work for the real data. We are exploring different sources of the shift.

Week 7 (3/15 and 3/17)

Explorations in the source of the echo have yielded little results. I am unhappy to report a lack of progress this week. Data collection and further checking of phase shifts within different components of the equipment are continuing.

Week 8 (3/22)

On Monday of this week we experimented with the equalizer of the data conditioning boxes. It seems that by turning off the equalizer the data collected is far less accurate, but we may remove the echo effect of the data! This will be explored further next week, but is great news.

Week 9 (3/29, 3/30, and 3/31)

This week we have confirmed that the echo effect is gone. We worked on applying triangulation methods to estimate the position of the source in three dimensions. Currently we are developing methods to project the direction of arrival onto the XY and YZ planes. We will use that data to estimate the actual position. It is working decently well at the moment, but different cases need to be set up for different angle ranges to ensure that our calculations fall within the domain of different trigonometric functions.

Files

Here are the files that we use. Note that the MATLAB files are in a .zip file because the wiki will not allow me to upload .m files.

File:AcousticVectorSensorMain.vi

File:Filter.vi

File:AVS Matlab Scripts.zip