臺灣博碩士論文加值系統

本論文探討救護車與消防車輛鳴笛聲之自動偵測方法。在一般道路上，救護車、消防車、或警車等執行任務時皆以警示鳴笛聲來提醒用路人禮讓，但由於近年來汽車隔音效果十分良好，或車內收聽音響等因素，常導致駕駛人未察覺警示鳴笛聲車輛而不知禮讓，或甚至在路口與警示鳴笛車輛碰撞。基於協助駕駛人察覺周遭任務車輛的考量，本論文嘗試開發警示鳴笛聲之自動偵測方法，初步僅以台灣地區的救護車與消防車為探討對象。我們將此偵測問題視為三種聲音的辨識問題，分別是救護車鳴笛聲、消防車鳴笛聲、與一般道路噪音。利用類神經網路學習這三種聲音的梅爾倒頻譜係數，並用以判斷每秒所錄下之未知聲音是屬於三種中的何者。經實驗測試，本論文所提出的系統在-15dB的模擬吵雜環境下約可達90%的警示鳴笛車種辨識率，而在道路上實測則約可達93%的辨識率。

This study investigates automatic methods for detecting ambulance and fire truck siren sounds. In general, emergency vehicles, such as ambulances, fire trucks, and police cars use sirens to warn other road users for quick passage, especially for moving through the traffic. However, due to the well sound insulation techniques in modern cars, drivers may not be aware of the approaching of the emergency vehicles, especially when in-vehicle audio systems are used. As a consequence, emergency vehicles may be blocked and even collided with other vehicles. To help drivers avoid being unaware of emergency vehicles, this work proposes automatic detection methods of siren sounds. But during the initial study stage, we focus only on the siren sounds of ambulance and fire truck in Taiwan. The detection task is formulated as a problem of identifying the three sound classes, respectively from ambulance, fire truck, and ambient noise. We propose a recurrent neural network with deep learning to determine which of the classes each 1-sec sound recording belongs to, based on its mel frequency cepstral coefficients. Our experiments shows that the proposed method can achieve the identification accuracy of 90% in simulated -15dB noisy sound data and 93% in real sound data recorded on a downtown street.

摘要 I
ABSTRACT II
誌謝 IV
目錄 V
表目錄 VII
圖目錄 IX
第一章 緒論 1
1.1介紹 1
1.2相關文獻探討 1
1.3問題定義 2
1.4論文架構 2
第二章 警示鳴笛聲訊號特性與研究方法 3
2.1警示鳴笛聲訊號特性 3
2.2研究方法 5
第三章 特徵擷取 6
3.1特徵參數擷取 6
3.1.1預強調(Pre-emphasis) 7
3.1.2音框化（Framing） 7
3.1.3漢明窗(Hamming Window) 8
3.1.4快速傅立葉轉換(Fast Fourier Transform) 9
3.1.5三角帶通濾波器(Triangular Bandpass Filter) 9
3.1.6離散餘弦轉換(Discrete Cosine Transform) 10
第四章　類神經網路 12
4.1簡介 12
4.2類神經網路特性[7] 13
4.3人工神經元 14
4.4激勵函數 15
4.5多層感知機(Multi Layer Perceptron) 16
4.6遞迴式類神經網路(Recurrent Neural Network) 20
4.6.1傳統RNN 21
4.6.2 LSTM-RNN 28
4.7 自應性學習率演算法Adam 37
4.8 Softmax 38
4.9 One-Hot Vector 39
4.10 Cross-Entropy 40
4.11使用架構 41
4.11.1架構一 多層感知機(Multi Layer Perceptron) 41
4.11.2架構二 LSTM-RNN 42
4.12類神經網路訓練與測試流程 43
第五章 實驗 45
5.1實驗資料與實驗環境 46
5.2警示鳴笛聲訊號識別實驗 49
5.2.1實驗一 多層感知機 50
5.2.2實驗二 LSTM-RNN 54
5.2.3實驗三 外部測試 59
第六章 結論與未來展望 61
6.1結論 61
6.2未來展望 61
參考文獻 62
附錄 64

[1].王文聖, 廖俊鑑, 連思斌, “以頻率特徵辨識救護車鳴笛聲” , International Journal of Advanced Information Technologies (資訊科技國際期刊), Vol. 6, No. 1, pp. 39-45, 2012。
[2].連思斌, 王文聖, 廖俊鑑, 「救護車鳴笛聲辨識方法」, 2011資訊科技國際研討會(2011 International Conference on Advanced Information Technologies, AIT), 287, Taichung, Taiwan, April 22-23, 2011。
[3].王文聖, 廖俊鑑, 彭聖竺, 「以應用於車載通訊之救護車鳴笛聲辨識方法」, 2013第七屆資訊科技國際研討會(2013 International Conference on Advanced Information Technologies, 2013 AIT), 169, Taichung, Taiwan, April 26-27, 2013K.
[4].Beritelli, F., Casale, S., Russo, A., andSerrano, S., “An Automatic EmergencySignal Recognition System for The HearingImpaired,” Proceedings of 12th DigitalSignal Processing Workshop and 4thSignal Processing Education Workshop,Jackson Lake Lodge Grand Teton NationalPark, Wyoming, U.S.A., pp. 179-182, Sept.
[5].2006.內政部所主管災害緊急應變警報訊號之種類、內容、樣式、方法及其發布時機公告，內政部，台內消字第1010821910 號，中華民國101 年4 月23 日。
[6].王小川, 語音訊號處理 ,臺北:全華圖書,2008
[7].張斐章、張麗秋, 類神經網路導論原理與應用, 臺中市:蒼海書局,2010
[8].蘇木春、張孝德, 機器學習-類神經網路、模糊系統以及基因演算法則,臺北:全華科技圖書有限公司,2002
[9].林昇甫、洪成安, 神經網路入門與圖形辨識, 臺北:全華科技圖書股份有限公司,2002
[10].S. Hochreiter and J. Schmidhuber. Long short-term memory. Neural Computation, 1997
[11].零基礎入門深度學習 https://zybuluo.com/hanbingtao/note/581764
[12].KINGMA, Diederik; BA, Jimmy. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014.
[13].黃致凱，利用心音進行身份識別，碩士論文，國立台北科技大學電子工程所，台北2016
[14].JOZEFOWICZ, Rafal; ZAREMBA, Wojciech; SUTSKEVER, Ilya. An empirical exploration of recurrent network architectures. In: Proceedings of the 32nd International Conference on Machine Learning (ICML-15). 2015. p. 2342-2350.
[15].ROSENBLATT, Frank. The perceptron: A probabilistic model for information storage and organization in the brain. Psychological review, 1958, 65.6: 386.
[16].HECHT-NIELSEN, Robert, et al. Theory of the backpropagation neural network. Neural Networks, 1988, 1.Supplement-1: 445-448.
[17].WERBOS, Paul J. Backpropagation through time: what it does and how to do it. Proceedings of the IEEE, 1990, 78.10: 1550-1560.