近年來隨著環保意識提升，電動車逐漸在全球盛行，特別是在台灣以電動機車為大宗。不過，電動機車雖然擁有低汙染、低噪音的優點，但是行人也容易因為它過低的音量而不易察覺有車輛靠近。生活中不乏聽到因為未注意到車輛行經而被嚇到的實例。台灣有非常多較小的巷弄是屬於行人與車輛共道，在沒有人行道的情況下，若是無法提前察覺到有車輛靠近，很容易因為反應不及而發生意外。人的聽力會隨著年齡退化，特別是對於高頻頻域。因此對於年長者來說，要及時閃避車輛是更加困難的。目前政府的對策為規定新款電動機車必須加裝低速警示，然而我們認為不只是在低速行駛的情況下才需要警示。我們想出若有一個偵測車輛靠近的裝置，能夠在有車輛靠近時藉由智慧型穿戴裝置發出警示給行人，如此一來便能改善因為聽不見車輛聲音而發生危險的情形。我們的研究主要為偵測電動機車的靠近，以及辨別它的來向。首先，我們用假人頭以雙耳錄音的方式蒐集了電動車從不同來向靠近的聲音，並建立資料庫。接著，我們將電動車所產生的特殊聲響以及雙耳線索作為特徵以隨機森林進行訓練。建立出來的模型能夠準確判斷出電動機車的靠近以及來向。此外，我們也訓練了將原始音檔加上加性高斯白雜訊後的模型來模擬在嘈雜環境中的偵測效果。最後，我們比較了使用隨機森林和支持向量機建立的模型，結果顯示隨機森林整體準確度較高，在誤報方面表現也較好。

Due to the rise of environmental awareness, electric vehicles have become popular all over the world. Electric motorcycle has become one of the most common vehicles in Taiwan. However, despite the advantage of low pollution and noise, pedestrians might neglect approaching electric vehicles since they are too quiet. We can often see examples of a person surprised by a vehicle that suddenly appears. In Taiwan, there are many alleyways with no sidewalks. If we are unaware of vehicles that are approaching in advance, there will be a high chance for accidents to occur. In general, people suffer from hearing loss with aging, and become less sensitive to high-pitched sound. As a result, it is more difficult for elders to dodge vehicles in a short time. The government made a policy that all the new electric vehicles should be equipped with a low speed alarm, yet we consider that this is inadequate for a pedestrian’s safety. We came up with an idea that there can be a vehicle detecting system, and pedestrians can get warnings sent it to their smart wearable devices. Our work is focused on the detection of an electric motorcycle and its coming directions.
First, we set up a dummy head with binaural recording to build a database of electric motorcycle’s sound. Next, we took the specific noise of an electric motorcycle and binaural information as features, and trained a random forest to detect that noise automatically. The model is robust on detecting an electric motorcycle and determining the direction of its approaching. Besides, we also used a dataset with additive white Gaussian noise to simulate the detection in a noisy environment. Finally, we compared the model built with random forest and support vector machine, and the results showed that random forest got a higher accuracy, and performed better in terms of reducing the false alarm rate.

[1] W. H. Organization, Hearing Loss Due to Recreational Exposure to Loud Sounds: A Review. World Health Organization, 2015.
[2] Z. Sun, G. Bebis, and R. Miller, “On-road vehicle detection: a review,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 5, p. 694–711, 2006.
[3] D. Dooley, B. McGinley, C. Hughes, L. Kilmartin, E. Jones, and M. Glavin, “A blind-zone detection method using a rear-mounted fisheye camera with combination of vehicle detection methods,” IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 1, p. 264–278, 2016.
[4] C. Chen and Y. Chen, “Real-time approaching vehicle detection in blind-spot area,” in 2009 12th International IEEE Conference on Intelligent Transportation Systems, p. 1–6, 2009.
[5] S. Smaldone, C. Tonde, V. K. Ananthanarayanan, A. Elgammal, and L. Iftode, “The cyber-physical bike: A step towards safer green transportation,” in Proceedings of the 12th Workshop on Mobile Computing Systems and Applications, HotMobile ’11, (New York, NY, USA), p. 56–61, Association for Computing Machinery, 2011.
[6] M. Takagi, K. Fujimoto, Y. Kawahara, and T. Asami, “Detecting hybrid and electric vehicles using a smartphone,” in Proceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing, UbiComp ’14, (New York, NY, USA), p. 267–275, Association for Computing Machinery, 2014.
[7] S. Kawanaka, Y. Kashimoto, A. Firouzian, Y. Arakawa, P. Pulli, and K. Yasumoto, “Approaching vehicle detection method with acoustic analysis using smartphone for elderly bicycle driver,” in 2017 Tenth International Conference on Mobile Computing and Ubiquitous Network (ICMU), p. 1–6, 2017.
[8] R. S. Woodworth and H. Schlosberg, Experimental Psychology. Oxford and IBH Publishing, 1954.
[9] D. S. Brungart and W. M. Rabinowitz, “Auditory localization of nearby sources. head-related transfer functions,” The Journal of the Acoustical Society of America, vol. 106, no. 3, p. 1465–1479, 1999.
[10] M. M. Van Wanrooij and A. J. Van Opstal, “Contribution of head shadow and pinna cues to chronic monaural sound localization,” J. Neurosci., vol. 24, no. 17, p. 4163–4171, 2004.
[11] H. Wallach, “On sound localization,” The Journal of the Acoustical Society of America, vol. 10, no. 4, p. 270–274, 1939.
[12] J. Hebrank and D. Wright, “Are two ears necessary for localization of sound sources on the median plane?,” The Journal of the Acoustical Society of America, vol. 56, no. 3, p. 935–938, 1974.
[13] A. Saxena and A. Y. Ng, “Learning sound location from a single microphone,” in 2009 IEEE International Conference on Robotics and Automation, p. 1737–1742, 2009.
[14] A. Ovcharenko, S. J. Cho, and U.-P. Chong, “Front-back confusion resolution in three-dimensional sound localization using databases built with a dummy head,” The Journal of the Acoustical Society of America, vol. 122, no. 1, p. 489–495, 2007.
[15] Y.-Y. Song and L. Ying, “Decision tree methods: applications for classification and prediction,” Shanghai Archives of Psychiatry, vol. 27, no. 2, p. 130, 2015.
[16] B. Gupta, A. Rawat, A. Jain, A. Arora, and N. Dhami, “Analysis of various decision tree algorithms for classification in data mining,” International Journal of Computer Applications, vol. 163, no. 8, p. 15–19, 2017.
[17] L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, p. 5–32, 2001.
[18] G. Biau and E. Scornet, “A random forest guided tour,” Test, vol. 25, no. 2, p. 197–227, 2016.
[19] T. M. Oshiro, P. S. Perez, and J. A. Baranauskas, “How many trees in a random forest?,” in Machine Learning and Data Mining in Pattern Recognition (P. Perner, ed.), (Berlin, Heidelberg), p. 154–168, Springer Berlin Heidelberg, 2012.
[20] C.-W. Hsu and C.-J. Lin, “A comparison of methods for multiclass support vector machines,” IEEE Transactions on Neural Networks, vol. 13, no. 2, p. 415–425, 2002.
[21] Y.-R. Chen, C. Y. Chang, and S. M. Kuo, “Active noise control and secondary path modeling algorithms for earphones,” in 2017 American Control Conference (ACC), p. 246–251, IEEE, 2017.
[22] G. Louppe, L. Wehenkel, A. Sutera, and P. Geurts, “Understanding variable importances in forests of randomized trees,” Advances in Neural Information Processing Systems, vol. 26, 2013.