- 學位論文
應用跨層最佳化與湧泉碼之即時無線影像串流設計與實現
Design and Implementation of Real-Time Wireless Video Streaming using Cross-Layer Optimization and Fountain Codes
摘要
為了要避免無線網路中重傳機制所造成的延遲效應，在無線串流系統中有大量的具有無碼率限制特性的湧泉碼設計被提出且廣泛應用。湧泉碼可以保留部分解碼過的資訊，並且一邊繼續解碼所接收的符元。當累積到超過預定的正確接收的符元量後，原本之資訊傳輸序列就可以完整地正確解出。我們在論文中提出一個使用了湧泉碼和可適性多進多出(multiple-input multiple-output) 正交分頻調變(orthogonal frequency division multiplexing) 無線網路技術之影像串流跨層最佳化演算法。當我們使用了可適性正交分頻調變技術再配合實際的通道狀態訊息，我們就可以調整傳輸之吞吐量來改善湧泉碼狀態下的無線通道效率。這個使用多進多出正交分頻調變無線網路技術之資源分配演算法可以讓所提出之無線影像串流系統與傳統的系統相比有8 dB的改善。 當考慮到其硬體實現的可行性時，當要實現湧泉碼中最先進的迅捷碼時會需要執行大維度的矩陣反轉，其維度會高達2^{16}。所以我們在論文中提供了一個基於索引機制之演算法來改善在線(online)迅捷碼解碼演算法。與使用相同在線解碼機制演算法的傳統迅捷碼相比，我們可以將演算法複雜度由矩陣維度之三次方減化到維度之二次方。而後我們使用了90 nm製造技術來實現了可重新配置之迅捷碼解碼器晶片。這顆晶片可以達到7 Mbps的吞吐量來達成SDTV 解析度下之壓縮影像無線傳輸且只需要52.7 mW 之功率耗能。此外，我們提出了另一個基於Sherman-Morrison定律之解碼演算法來達成非在線(offline)之迅捷碼解碼演算法。與一般的在線迅捷碼解碼演算法相比，其複雜度只需要原先之6.5 %。基於此演算法，我們實現了高效能之硬體設計且使用了Xilinx Kintex-7 FPGA板來驗證其設計。其吞吐量可以達到 46.5 Mbps，為使用前演算法之硬體設計的兩倍，且耗能沒有增加。 最後，我們提出了一個包含低複雜度湧泉碼解碼器之即時無線影像串流的原型設計展示。配合此高效能之湧泉碼解碼器，我們所提出使用FPGA實現之迅捷碼解碼器可以完全符合下一世代之無線影線串流系統之規格需求。因此，我們使用FPGA 來實現了迅捷碼和多進多出正交分頻調變無線解碼器進而達到即時之無線影像串流系統；搭配了RF 前端元件和軟體實現之影像編解碼器，我們完成了即時無線影像串流原型系統之設計、實現、以及驗證。
並列摘要
In order to avoid retransmission latency, Fountain codes, also known as rateless codes, have been widely proposed and applied in the designs for streaming systems. Fountain codes retain the partially decoded information, and continue to receive and decode the coded symbols until the number of accumulated correctly received coded symbols exceeds a pre-determined threshold, and then the complete information sequence can be recovered. This thesis presents a novel solution to cross-layer optimization for fountain code-based video streaming over adaptive multiple-input multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM) wireless networks. Armed with the adaptive OFDM in combination with the practical channel state information, the transmission throughput can be adjusted in order to improve channel efficiency when fountain codes are used. A resource allocation algorithm is proposed for wireless video streaming using MIMO-OFDM that achieves an improvement in SNR of 8 dB over the conventional approach. When considering the feasibility of hardware implementation, RaptorQ code, the most advanced solution for fountain code, requires that a huge matrix inversion be performed, typically in a dimension of up to 2^{16}. This thesis offers an index-based algorithm that is helpful in improving the online decoding of the RaptorQ decodingii algorithm, employing the same decoding mechanism as the conventional RaptorQ, but reducing the computational complexity from cubic to quadratic. Then, an IC implementation of a configurable RaptorQ decoder in 90 nm technology is presented. This chip achieves an average throughput of 7 Mbps at a power consumption of only 52.7 mW, while supporting compressed video streaming at a quality of up to SDTV (720x480). Furthermore, a low-complexity RaptorQ decoding algorithm is proposed that uses the Sherman-Morrison formula to achieve offine decoding of the RaptorQ decoding algorithm. Compared to the online RaptorQ decoding algorithm, the complexity of the proposed RaptorQ decoding algorithm is only 6.5%. Thus, based on this algorithm, a high performance hardware implementation is designed, and validated using a Xilinx Kintex-7 FPGA board. The throughput of the proposed design reaches 46.5 Mbps, which is more than twice that of previous works, while consuming almost the same power. The thesis finally provides a prototype for how a real-time wireless video streaming system that includes a low latency RaptorQ code will function. Exploiting the high-performance RaptorQ decoder, the FPGA implementation for the proposed RaptorQ decoder is able to fully meet the requirements for the next-generation of wireless video streaming systems. Hence, an FPGA implementation of both a RaptorQ decoder and a MIMO-OFDM receiver for a real-time wireless video streaming system is presented. Together with an RF front-end and a progressive video codec implemented via software, the design, implementation, and validation of a prototype for a high-performance real-time wireless video streaming system are accomplished.