Master of Science (MS)
Compression plays a significant role in Body Sensor Networks (BSN) data since the sensors in BSNs have limited battery power and memory. Also, data needs to be transmitted fast and in a lossless manner to provide near real-time feedback. The paper evaluates lossless data compression algorithms like Run Length Encoding (RLE), Lempel Zev Welch (LZW) and Huffman on data from wearable devices and compares them in terms of Compression Ratio, Compression Factor, Savings Percentage and Compression Time. It also evaluates a data deduplication technique used for Low Bandwidth File Systems (LBFS) named Two Thresholds Two Divisors (TTTD) algorithm to determine if it could be used for BSN data. By changing the parameters and running the algorithm multiple times on the data, it arrives at a set of values that give >50 compression ratio on BSN data. This is the first value of the paper. Based on these performance evaluation results of TTTD and various classical compression algorithms, it proposes a technique to combine multiple algorithms in sequence. Upon comparison of the performance, it has been found that the new algorithm, TTTD-H, which does TTTD and Huffman in sequence, improves the Savings Percentage by 23 percent over TTTD, and 31 percent over Huffman when executed independently. Compression Factor improved by 142 percent over TTTD, 52 percent over LZW, 178 percent over Huffman for a file of 3.5 MB. These significant results are the second important value of the project.
Raju, Robinson, "COMPRESSION OF WEARABLE BODY SENSOR NETWORK DATA USING IMPROVED TWO-THRESHOLD-TWO-DIVISOR DATA CHUNKING ALGORITHM" (2018). Master's Projects. 613.