Parallel Programming for the Implementation of RRO-NRDPWT-based ECG Compression System on Multi-Core Platform

[[abstract]]多核心的處理器架構，已成為現今電腦和智慧型行動裝置的主流，然而若要完全發揮多核心處理器的效能，除了硬體和作業系統的支援外，程式更要以平行編程的技巧來設計編寫，否則程式的運作將依舊被侷限在單一核心上。 而在行動裝置普及後，一般預料穿戴裝置與物聯網的相關應用，將為下波產業成長注入新的動能，其中有關智慧醫療的相關產品，更深具發展的潛力。目前已有廠商研發出可攜式的穿戴型心電圖量測儀器，若要更進一步與物聯網和雲端概念結合，都需藉由網路來傳遞心電圖資訊，由於完整的心電圖量測包含了12個導極，為了因應長時間的量測需求，所需處理的資料量也勢必相當龐大，若能透過良好的壓縮演算法，先將心電圖資料壓縮後再予以傳輸，不但能達到節省頻寬的效果，更能縮短傳輸的時間進而達成同步傳輸的目的。 因此，本論文以實驗室所研發的「可逆式四捨五入非遞迴離散小波轉換」心電圖壓縮系統為理論基礎，搭配上時下主流的多核心智慧型行動裝置，以自行歸納出的平行運算編程技巧，完整實作出涵蓋小波轉換、量化和編碼的心電圖壓縮系統。最後，在本論文所使用的四核心開發平台上，成功的將壓縮時間之加速比(Speedup)提昇達3以上。[[abstract]]Nowadays Multi-Core processors have become the mainstream architecture of PC and smart mobile devices However if we want to take full advantage of Multi-Core architecture in addition to the support of hardware and operating systems programs should be designed and implemented with parallel programming techniques; otherwise the execution of a program will still be limited to single core Due to the popularity of smart mobile devices it is generally expected that the relevant applications of wearable devices and IOT will be contributing to the growth of the technology industry Among these applications the intelligent medical products are potentially huge There are some manufacturers that have already developed wearable ECG monitoring devices by now If we want to take it one step further by integrating the concept of IOT and Cloud the ECG data needs to be transmitted over the network However because an ECG monitoring device consists of 12 leads in order to meet the requirement of long term monitoring it must be able to handle large data sets If we can use a good compression algorithm to compress the ECG data before transmission it will not only save bandwidth but also reduce the transmission time to achieve the purpose of real-time transmission This thesis is based on the theorem of a RRO-NRDPWT-based ECG compression system proposed by our lab which consists of a wavelet transform quantization and coding module In this thesis I also summarize some parallel programming techniques to implement the complete ECG compression system on the mainstream Multi-Core platform In conclusion the execution time of compression on a Quad-Core device is successfully improved and the speedup is more than