本論文之研究應用於電子化工廠無線感測網路,重點在於如何將無失真資料壓縮演算法實現在電子化工廠無線感測網路上,研究主要以感測工廠內機械轉動的聲音訊號為主,而感測的目的是為了判斷工廠內的機械轉動聲音是否為正常運作。在開發階段使用個人電腦為平台配以DAQ設備來擷取資料,其感測訊號源以機器轉動時所發出的音源,並且在感測端上透過C/C++程式做擷取與即時的壓縮演算法,在透過無線感測網路將壓縮資料傳送至中控端。然後在中控端會透過我們所設計的解壓縮演算法來還原回原始資料讓中控端來做使用。本文主要以研究兩種不同的無失真資料壓縮演算法實際應用於壓縮系統中,並且在同樣的聲音訊號來源下複雜度與壓縮率之比較,在設計的方面主體由正規化最小平均平方(Normalized least mean square)演算法預測模型與串接霍夫曼編碼跟萊斯碼等熵編碼組成。
This thesis is a research about wireless sensor networks in e-factories. The emphasis of this study is on how to apply lossless data compression algorithms on wireless sensor networks in e-factories. This research focuses primarily on detecting and processing audio signals generated by the rotation of machines in the factories in order to determine if machines are operating normally. In the development phase, a personal computer equipped with a data acquisition (DAQ) device is used as the platform to capture data. The source of the sensing signal is the sound emitted by machine while it is running. At the sensing terminal, programs written in C/C++ performs data capture and real-time compression algorithms; compressed data is then transmitted to a central control terminal through the wireless sensor network. At the central control terminal, data is reverted back to its original format using self-developed decompression algorithms for further processes. This study mainly investigates the actual applications of two different types of lossless data compression algorithms in compression systems. Comparisons of degrees of complexity and compression ratios between the two algorithms are made using the same audio signal source. The design of the compression algorithm employs normalized least mean square (NLMS) algorithm for prediction, and entropy codes including concatenated huffman codes and Rice Code.