I graduated from Faculty of Electrical Engineering and Computing ( FER ) at Zagreb in the field of computer science. During my studies I worked on many small projects at FER ranging from creating Java applets used in class up to my thesis on compression in embedded devices. Here is a short description of some of my projects.

Thesis

Subject of my graduation thesis was compression on wireless sensor nodes using wavelets. It was aimed at Mica2 wireless sensor nodes ( or motes ). These nodes are much more limited in computing resources if compared with PDAs or personal computers. They are designed mainly to last as long as possible given their only source of energy are a pair of AA batteries. Mica2 motes can be fitted with sensors ranging from microphones, humidity sensors to accelerometers.

Mica2 motes use a special operating system called TinyOS. It uses events rather than threads for multitasking. A subset of the programming language C called NesC is used to create software running on TinyOS.

Aim of my thesis was to see if compression implemented on these small devices could yield significant energy savings and prolong life of the these small devices. Compression had to be adapted to measuring vibration. I intended to adapt it by implementing compression using the wavelet transform.

Wavelet transform could be described as a way of transforming a signal from the time domain into a time-frequency domain. The result of a wavelet transform gives both information about what frequencies are present in a signal and when they appear. A very good thing that goes in favor of using a wavelet transform is that it can be implemented very efficiently and only using integer operations. Mica2 motes have no floating point hardware and working with floating point operations is very slow.  Another big plus for the wavelet transform was that it made implementing lossy compression possible.

My thesis showed that it is possible to efficiently implement compression using wavelet transform on Mica2 motes. Lossless compression turned out not to improve very much with use of the wavelet transform. However when doing lossy compression wavelet transform made it possible to leave importan frequency bands intact while reducing or removing those bands that either contained less energy or were considered unimportant.