透過您的圖書館登入
IP:3.19.56.114
  • 學位論文

應用支持向量機於鯨豚哨音分類之研究

The Analysis of Identification of Cetaceans’ Whistle With Support Vector Machine

指導教授 : 黃乾綱

摘要


利用鯨豚哨音特徵去識別鯨豚種類,不論是台灣或者國外研究大多是使用鯨豚哨音在時頻圖(spectrogram)上的特徵為基礎去發展,並且輔以專業的人工辨識(ECO,experienced human operators ),藉由聽取聲音樣本的同時,再以視覺比對訊號的時頻圖(spectrogram),但由於不同種類鯨豚的哨音其外型都很類似,無法由肉眼直接判斷哨音為哪一種鯨豚的哨音,所以本研究採用的是一種以模擬人耳聽覺特性的方式,去找出鯨豚哨音的特徵,基於聽覺知覺特性的特徵擷取技術,我們比較廣泛常見的是梅爾倒頻譜係數(Mel-frequency Cepstral Coefficients, MFCC),它主要原理是模擬了人耳聽覺上的二個效應:遮蔽效應(Masking Effect)和臨界頻帶(Critical Band),然後去萃取語音訊號的特徵值。 最後,分類的方式我們採用的是支持向量機(SVM),所謂支持向量機(SVM)是一種由Vapnik等根據統計學習理論提出的一種新的機器學習方法。主要用在處理分類上的問題,支持向量機(SVM)在解決非線性及高維模式識別問題中表現出許多特有的優勢,簡單來說,對於一群資料而言,有時候我們會希望依據資料的一些特性來將這群資料分為兩群,支持向量機(SVM)的概念就是首先將資料投影至高維度的特徵空間中,再去找出將兩個不同的集合分開的超平面(hyperplane),進而達到分類的目的。 初步的實驗結果顯示出本論文所提出的作法對於鯨豚哨音辨識率有相當顯著的提昇,期望之後能再次提升辨識度,對於未來鯨豚哨音研究將是一項不可或缺的利器。

並列摘要


Whales and dolphins use the whistle to identify cetacean species characteristics, whether in Taiwan or abroad study mostly cetaceans whistle when using frequency map (spectrogram) on the basis of characteristics to develop, and supplemented by professional artificial identification. By listening to the sound samples, then compared the visual signal when the frequency diagram (spectrogram). However, due to different types of whales and dolphins whistle which looks very similar, can not be directly judged by the naked eye whistle to whistle which cetacean. Therefore, this study is a way to simulate human auditory characteristics to identify the characteristics of cetacean whistle, capture technology is based on the characteristics of auditory perceptual characteristics Finally, the classification of the way we use the support vector machine (SVM). Support vector machine (SVM) is a new kind of machine learning method based on statistical learning theory Vapnik proposed Preliminary experimental results show that this paper proposed approach for the identification of whales and dolphins whistle quite a significant rate increase, after hoping to enhance recognizable again, whistle cetacean research for the future will be an indispensable tool

參考文獻


2. 王瑋強 and W.-C. Wang, 應用支持向量機於判斷中文停用字之研究
5. 黃禎智, 基於語者模型梅爾倒頻譜係數與類神經網路之主從式即時語者辨識系統. 2005.
1. Alpaydın, E., Introduction to Machine Learning. 2010.
The Analysis of identification of Chinese Stop Characters with Support Vector Machine.
3. 林宗勳, .

延伸閱讀