- 學位論文
在藍芽環境中傳送H.263視訊資料之錯誤偵測
Error Detection of H.263 Video Data Transmitted in Bluetooth Environment
摘要
藍芽(Bluetooth)是當今短距離無線通訊的主要標準之一，利用藍芽模組來傳送視訊資料將可運用於PDA、筆記型電腦以及其他含有顯示裝置的無線傳輸設備中。H.263視訊壓縮標準，主要是應用於低頻寬頻帶上視訊資料的傳輸，符合藍芽的傳輸速率，另外其提供的畫面大小也恰合適應用於上述的裝置中。 本論文將以藍芽封包格式來傳輸視訊資料，在遭受高斯雜訊及Rayleigh Fading干擾時，探討在不同視訊品質下的錯誤偵測問題。視訊資料在傳輸前先經H.263的視訊壓縮，再經過藍芽封裝及調變，之後傳送至無線環境中。在接收端則會經藍芽及H.263的視訊解碼端解回視訊畫面，再經由本論文所提出的錯誤偵測法尋找出錯誤的區域。好的錯誤偵測結果對其後的錯誤遮隱以及視訊品質的提升有莫大的效益。 模擬顯示，使用有或無保護的藍芽封包以及不同的信號對雜訊比值(Signal to Noise Ratio, SNR)，視訊畫面會受到不同程度的干擾。最後採用錯誤偵測(Error Detection)的技術，把錯誤的 方塊找出來。結果顯示即使是背景複雜的Salesman畫面，在SNR = 16dB時，平均錯誤偵測率(Detection Rate, DR)仍可達85%以上。還有，所有錯誤偵測的平均的錯誤率(False Detection Rate, FDR)都可壓抑在15%以下。
並列摘要
Nowdays Bluetooth (BT) is one of the major standards for short-range wireless communications. Video transmission techniques with BT modules can be used in PDA, notebooks, and other wireless equipments with displaying devices. The H.263 video compression standard is mainly for the transmission of video data in low-bandwidth channels. It fits the BT transmission rate. In addition, the provided frame sizes are suitable for the equipments described above. In this thesis we discuss the error detection problem for different quality of video transmitted in BT packet formats under the AWGN noise and Rayleigh fading. The video data are compressed first using the H.263 compression technique before they are transmitted. Then they are packetized in BT formats and are modulated before transmitted in a wireless environment. In the decoding process at the receiving end, the video will be decoded through BT and H.263, and the error region will be detected using the proposed error detection method in this thesis. Good error detection results greatly facilitate the subsequent error concealment step and the improvement of video quality. Simulation results show that using error protected/unprotected BT packets and different signal to noise ratios will result in different levels of video degradation. Finally, we use the error detection technique to locate the blocks in error. The results show that the average error detection rate can still maintain more than 85% for the background-complicated Salesman sequence when SNR = 16dB. Furthermore, all the average false detection rates are kept below 15%.