6.1 結論
現行的車輛偵測方法主要透過訊號能量的大小作為設定門檻基準，傳統計算訊號 能量的方法是以每一個時間點(幀)上的頻譜能量總和作為每幀的能量值，以這種能量 計算方法得到的門檻值，在目標(車輛)訊號與雜訊的能量相差大時，可以辨識出車輛 訊號，但在實際車輛偵測時候，常會有目標訊號能量與雜訊相差不大的情況。為了在 這種情況下，系統仍有準確的車輛偵測能力，本研究使用譜熵方法來計算訊號能量，
由此決定門檻設定機制與端點檢測模式進行車輛偵測。本研究成果整理如下：
(1) 在模擬訊號實驗中，不論在哪一個信噪度下，本研究所提出之方法與傳統能量門 檻法相比，其訊號辨識結果皆較為準確，且在越低信噪度的環境，兩個方法的準 確度差異越大。
(2) 以雷達偵測實際收取資料來進行訊號偵測實驗，由實驗結果發現，距離雷達偵測 器較近的車道(如第二、三與四車道)，本研究的譜熵門檻辨識結果與能量門檻辨 識結果相比，本研究之方法分別以 97.50%、93.33%、87.10%略優於能量門檻方 法，辨識準確率差約在 5%以內。
(3) 而距離較遠的車道(如第五、六車道)，根據實驗結果可以明顯看到本研究的辨識 結果優於能量門檻辨識結果，其辨識準確率分別差 10.00%與 14.29%。
(4) 推測本研究之譜熵門檻法因為採用資訊熵的概念，在計算每一幀的訊號能量時，
同時包含了該幀上的譜熵分布與整個幀的能量大小，相較於傳統能量門檻單純考 慮幀上的能量總和，本研究方法擁有較多的訊號資訊，在目標訊號與雜訊差異小 的情況下，本研究方法提供車輛偵測的資訊較多，因此準確度也較高。
6.2 建議
本研究建議整理如下：
(1) 由於目前沒有論文以譜熵門檻檢測法應用於雷達偵測器的車輛有無辨識，故本研 究目前對於參數的設定，如α、β值，主要以使用的雷達偵測器訊號特性做設定，
未來可以進一步研究，如使用訊號誤判率或最小平方法來決定這些參數。
(2) 本研究目前沒有把帶通濾波器設計納入演算法中，因此在後續研究若加入帶通濾 波器設計，可以針對雷達偵測器的特性，設計合適的濾波器函數，有助於增加偵 測器的辨識能力。
51
(3) 對於背景訊號的設定，是以無車階段的學習期間，以第一幀作為背景訊號值，經 由訊號處理過程予以扣除，後續研究可以根據端點檢測的結果來對於針對背景訊 號進行更新，以更符合道路情形變化。
(4) 以本研究方法可以更準確地判斷出有無車輛通過雷達偵測範圍，且根據實驗過程 可以發現大、小型車輛的譜熵有所差異，未來可針對車種進行辨識，更進一步提 升系統辨識能力。
52
參考文獻
[1] Bullington, K. and J. M., Fraser, "Engineering aspects of TASI", Bell System Technical Journal, pp.353-364, March 1959.
[2] EIS Inc., "RTMS User Manual", EIS Electronic Integrated Systems Inc., 2002.
[3] Fast , J. D.," Entropy, the significance of the concept of entropy and its application in science and technology", Macmillan,1982.
[4] Genderen, P., "Feature based plot classification using a Bayes algorithm", Proceedings of the 3rd European Radar Conference, pp. 41-44, 2006.
[5] Guo, Q., N. Li, and G. Ji, "A improved dual-threshold speech endpoint detection algorithm", ICCAE Computer and Automation Engineering, 2010.
[6] Haigh, J. A. and J. S. Mason, "Robust voice activity detection using cepstral features", IEEE Computer, Communication, Control and Power Engineering conference, 1993.
[7] Hussain, A., S.A. Samad and L.B. Fath, "Endpoint detection of speech signal using Neural Network", IEEE Trans on ASSP, pp.271-274, 2000.
[8] Jaynes , E.T., Information Theory and Statistical Mechanics II, Physical review, 1957.
[9] Jou, Y.J. and Y.K. Chen, "The online vehicle type classifier design for road-side radar detectors", AIP Conference Proceedings, vol. 1148, Issue 1, pp.462-465, 2009.
[10] Junqua, J.C., M. Brian and R. Ben, "A robust algorithm for word boundary detection in the presence of noise", IEEE Transactions on Speech and Audio Processing, vol. 2, no. 3, 1994.
[11] Lamel, F., R. Rabiner and E. Rosenberg, "An improved endpoint detector for isolated word recognition", IEEE Transactions on Acoustics, Speech and Signal Processing, vol. 29, no. 4, 1981.
[12] Leduc, G., "Road traffic data collection methods and application", Working papers on Energy, Transport and Climate Change N.1, 2008.
[13] Leung, S. W. and J. W. Minett, "A fuzzy shadow feature scheme for radar signal detection", International Conference of Information, Communication and Signal Processing, vol. 3, pp.1386-1388, 1997.
[14] Li, Y., Y. Fan, "A novel algorithm to robust speech endpoint detection in noisy environments", IEEE Industrial Electronics and Application, 2007.
53
[15] Li , Q., J. Zheng and A. Tsai, "Robust endpoint detection and energy normalization for real-time speech and speaker recognition," IEEE Transactions on speech and audio processing, vol.10, No.6,March 2002.
[16] Lin, C.T., G.D. Wu, "A robust word boundary detection algorithm for variable noise-level environment in cars", IEEE Transactions on Intelligent Transportation Systems, vol.3, No.1, March 2002.
[17] Malboubi, M., J. Akhlaghi, and M. Saraf, " The intelligent identification of air and sea targets in coastal radars", Proceedings of the 3rd European Radar Conference, pp.
17-20, 2006.
[18] McClellan , J. H., R. W. Schafer and M. A. Yoder, Signal Processing First, Pearson Prentice Hall, 2003.
[19] Musch, T., "A high precision 24 GHz FMCW-radar using a phase-slope signal
processing algorithm", Proceedings of the 32rd European Microwave Conference, pp.
1-4,2002.
[20] Ney, H., "An optimization algorithm for determining the endpoints of isolated utterances", IEEE Transactions on Acoustics, Speech and Signal Processing, vol. 6, pp.720-723, 1981.
[21] Rabiner, R. and R. Sambur, "Voiced-unvoiced-silence detection using the Itakura LPC distance measure", IEEE Transactions on Acoustics, Speech and Signal Processing, vol. 2, pp.323-326, 1997.
[22] Shannon, C.E., "A mathematical theory of communication", University of Illinois Press, Urbana, 1949.
[23] Shen, J.L., J.W. Hung and L.S. Lee, "Robust entropy-based endpoint detection for speech recognition in noisy environments", presented at the ICSLP, 1998.
[24] Stove, A.G., "Linear FMCW radar techniques", IEEE Proceedings of Radar and Signal Processing, vol. 139, no. 5, pp. 343-350, 1992.
[25] Wu, B.F. and K. C. Wang, "Robust endpoint detection algorithm based on the adaptive band-partitioning spectral entropy in adverse environments", IEEE Transactions on speech and audio processing, vol.13, No.5,March 2005.
[26] Wu , G.D. and C.T. Lin, "Word boundary detection with mel-scale frequency bank in noise environment", IEEE Transactions on Speech and Audio Processing, vol. 8, no.
5, 2000.
54
[27] Zhang, H., W. Yu and X. Sun, "A novel method for background suppression in millimeter-wave traffic radar sensor", Proceeding of the 11th International IEEE Conference in Intelligent Transportation Systems, pp. 699-704, 2008.
[28] Zhang, S., J. Fan, L. Shou and J. Dong, "A detection method of radar signal by wavelet transforms", Fuzzy Systems and Knowledge Discovery Forth International Conference, vol. 2, pp.710-714, 2007.
[29] 交通部統計處 - 97年重要交通統計指標分析，2009。
[30] 林大傑，車輛偵測器不完整資訊推估旅行時間之研究，逢甲大學交通工程與管 理所，碩士論文，民96。
[31] 卓訓榮等，「自動化路況資訊偵測系統研發與示範(一)」，交通部運輸研究所 委託，民96。