海豚哨叫聲偵測之研究

Abstract

臺灣於彰化至苗栗將設置離岸風場，由於該風場緊鄰中華白海豚生態保護區，於未來風場營運時建立一套對鯨豚之被動式聲學監測系統為發展要務。本研究於2017年4月至7月於苗栗竹南海域佈放五處水下被動式聲學測站，並將回收音檔以Matlab 2015與Python所撰寫之非針對性哨聲偵測程式加以偵測分析；發現於4~5月中華白海豚族群活動較為接近苗栗竹南海域，五個水下測站共偵測到599聲哨叫聲。七月所偵測到之哨叫聲極少，僅P3與P5點位有零星偵測；整體而言本研究所偵測到哨叫聲數量偏少、分布集中；以偵測效能而論，偵測程式對於〖SPL〗_rms大於 80 dB且SNR大於10dB之哨叫聲其偵測率達90%以上，並擁有12%以下誤報率。 進一步檢視本程式所偵測之哨叫聲時點，發現於多台水下麥克風測站都有偵測到同一聲哨叫聲，計算其音傳損耗可推估海豚發聲聲源之位置，達成粗略之聲源定位。為了驗證哨聲偵測器是否能對海豚聲源進行更準確定位，本研究於高雄港進行人造升頻聲源模擬實驗，利用到達時間差(Time Difference of Arrival)對第一圈及第二圈測線聲源進行定位，定位誤差平均值約50米。總體而言；本研究為被動式聲學的偵測與定位層面之進一步應用，於未來臺灣西海岸被動式聲學監測網建立提供關鍵技術之初始參考方向。

This study provides a feasible, practical and theoretical basis for the plan of Passive-Acoustic-Monitoring System for Chinese White Dolphins. Five Acoustic under water recorders were placed at the shallow water near Miaoli, Taiwan. To extract the whistles from the audio files, the Whistle-detector algorithm was used to recognize the patterns of whistles which shows on the spectrums from raw data . The result of detection indicates that there were 599 times of whistles recorded by hydrophone from April to May, and barely detected in July. In order to put Whistle-detector into the practical applications, the experiments to localize sound source were done near the Kaohsiung Harbor. The array, which consists of three hydrophones, were used to record the chirp signals from Marine III research ship. Each recorder received the chirp signals at the different time. Therefore, time difference of arrivals (TDOA) was the suitable way to compute the localization of sound sources. The average of the positioning errors was 50 m. According to the result of the TDOA experiment and the performance of the Whistle-detector, a more comprehensive investment is needed for the passive acoustic monitoring (PAM) in the future.