Fourth month at Internship

The agitation of the previous weeks work hadn’t abated at the beginning of the week. As the code maintainer had arrived from the annual leave, I had discussion with him about the code flow and did some final touch ups.

An interesting event of the month was the New Year celebrations. It started with conveying the wishes and small talk about new year, and of course Sri Lankan economics and politics – Mervin and Bro’s Ltd hadn’t had a show then though, moved into traditional part, lighting the auspicious lamp, followed by the speech by General Manager Anuradha Tennakon, culminated with award ceremony and concluded with new year dishes being served. In his speech, GM referred to the shift to new building, Motorola’s acquisition and remarkable success of the company in the last year with about 25 patent disclosures have been accepted.

Also a full test on UCA, the device for which we had spent hours on adding extension to power stat, as its tiny display didn’t allow us to perform anything more than seeing the title bar of the application, was conducted in this month. Following the discussion on the test results, a peculiar behaviour under discharging was observed, and it required us to test the discharging again at higher resolution, which at the end revealed an interesting fact.

As samples were collected at 30 second interval, sudden surge of current happening every 16 seconds due to the activity of radio module, upon the accumulation, produced a strange pattern at 7 minute interval.

Also completed the Powerstat application, and uploaded to the repository and archived this week. As the application is complex with versatile use, a user document was written where simple info page format was followed.

Another interesting event of the month was First Aid Program conducted by Red cross, Sri Lanka. We were given introduction into CPR and practical demonstration. Further the instructor led us into many case studies and discussed how they could have been handled differently. Further, necessary actions to be taken in the event of burns, fractures and other critical situations were discussed. Also we were encouraged to delve into the topic and rationally opt the action.

As the project I had been working on reached the completion, I got an R&D project -palm print recognition and its potential use as user identification / authentication. Although it is currently only in primitive state, with a patent application filed and not many studies are carried out in the area, its potential use has been widely acknowledged. Search for possible solutions for palm print sensing algorithms from internet was, hence, fruitless. And, it was decided to try correlation and find the common patterns and layouts of the palm prints. We took samples of palm prints with the high contrast imager that was available in MC 70 devices. However, later for convenience, it was decided to use standalone imagers.

In correlation, two patterns are correlated in 2-D plane, and if the result is above a threshold value it is accepted. However when two images are not equally scaled or properly arranged correlation coefficient drops significantly thus invalidating the use of the method. We could, however, pre scale the images, and rotate one with respect to another, and correlate, thus overcome this issue. Nevertheless the processing power required and time delays would become a huge concern especially when this is deployed in an embedded device such as a mobile computer.

I was also asked to look into the potential of using camera module which is usually available in phones as a tracking device similar to optical mice. Referred the optical mouse specifications and learnt, to track a motion of 10 cm/s with a 32x32 pixel frame we will have to sample at a rate greater than 160 samples/s. However for a smooth, fault free tracking when noise is present the sampling requirement rises upto 1000 samples/s, where in commercial optical mice it is uncommon to see a sampling rate of 6000 samples/s. No sooner, I thought it would be impossible to use a camera module as it operates at merely 24 samples/s, found there has been a research done in china, sponsored by Nokia, which succeeded in using Nokia 6630 device to track. It takes samples at 25 samples/s rate and from a single frame (1024x1024 pixels) several sample frames (around 40) were obtained, thus a fair rate of 1000 samples/s was achieved. However in our case, as it has already been published the idea was dropped.

So I am looking forward to continue my research in Palm print recognition algorithm, which would require extraordinary image processing and noise reduction as only a 640x480 pixel intensity based imager is used to acquire pictures and palm doesn't contain clear ridges and valleys as finger prints.