G.729A快速搜尋演算法整合JPEG影像壓縮技術實現於監視系統

Implementation of JPEG and G.729A Fast Search Algorithm on an Internet Monitoring System

本文整合語音與影響壓縮編碼技術與網路通訊協定實現一個以PC 為平台且具有人機介面多功能視窗應用程式監視系統。本系統具有(1)即時監視；(2) 即時偵測監視區域異動與自動發簡訊及Email；(3)透過網路直接進入系統主機進行遠端即時監視；(4)即時影像電話與監視區域進行對話；以及(5)多媒體播放器。當監視區域平均圖素變動超過一設定臨界值時系統即自動發簡訊及Email 通知管理人員，且所拍的影像以JPEG 壓縮演算法進行壓縮儲存並進行遠端備份，另外對於語音訊號之處理，是採用ITU-TG.729A 語音壓縮編碼演算法，如此可以大量減少硬碟儲存所需的容量及網路傳輸所需的頻寬，遠端備份被壓縮的影像及語音訊號有利於資料之保存。為了減少G.729A 語音壓縮編碼計算複雜度，本文提出一快速搜尋演算法來改善ACELP(Algebraic Code Excited Linear Prediction)激發碼簿搜尋之效率，其技術是利用一設計的函數來預估篩選每一子音框之候選的激發脈衝位置，以進一步降低ACELP 碼簿搜尋激發脈衝所需之計算量。經實驗結果顯示本文提出的快速搜尋演算法相較於原本G.729A 語音編碼器可改善編碼效率約70%至21.25%的相對計算量，而語音品質PESQ (Perceptual Evaluation of Speech Quality)平均僅少許下降約0.017 至0.002，且人類絲毫感覺不出語音品質下降，客觀評估證明本文提出的快速搜尋演算法所產生的語音幾乎接近於原來編碼器輸出的語音品質。

This paper implemented an internet monitoring system on a PC with window graphical user interface by employing image compressing (JPEG), speech codec (G.729) and network protocol techniques. This system has the following functions, namely, (1)real time monitoring;(2)real time motion detection and automatic short message service and send Email;(3)remote internet monitoring;(4)real time contact with monitoring areas; (5)multimedia players. When the average variation of image pixel in the monitoring area exceeds the threshold value, the system automatically send short message service and email for advising administrator. The captured image was compressed by the JPEG algorithm to backup for local and remote server. In addition, the speech signal was compressed by the G.729A speech codec and the processed data can reduce the hard disk storage capacity required and bandwidth required for network dramatically. The remote backup preserves evidence helpfully. To reduce the computational complexity of ACELP (Algebraic Code Excited Linear Prediction) in ITU-T G.729A speech codec, in this paper, we propose a fast search algorithm which uses a designed function to predict the candidate pulses for each subframe. Simulation results show that the average of the perceptual evaluation of speech quality (PESQ) score is degraded slightly, by only 0.017 to 0.002, and our proposed method can reduce computational complexity by 70% to 21.25% relative to the original G.729A coding computation load with perceptually negligible degradation. Our objective evaluations verify that the proposed fast search scheme can provide speech quality comparable to the original codec approach.