Localization of Bryde's whales using time difference of arrival principles

Cormick, Joshua (2021-03)

Thesis (MEng)--Stellenbosch University, 2021.

Thesis

ENGLISH ABSTRACT: The thesis presented gives the detailed practical implementation for a passive audio de-tection system that makes use of time difference of arrival theory to locate the source ofaudio signals emitted from Bryde’s whales. The system consists of three individual float-ing sensors deployed in the water that communicate with the user-unit on the researchvessel. The sensors are comprised of Raspberry Pis and various other specific hardware.The transfer of the sensor readings and the calculations done with said data allows for thelocation of the audio signal to be calculated through a time difference of arrival algorithm.The overall audio signal source localization system is made possible by integrating varioussub-systems. These being; detection, sensor tracking, communication, time synchroniza-tion, and signal location calculation. The theoretical principles behind each of thesesubsystems are discussed as well as their practical implementation, simulations and test-ing. In order to locate the source of an audio signal, one must first be able to identify whetheran input signal to the system is the desired signal. A dynamic time warping algorithm ismade use of in order to detect the desired audio signals. The validity of this algorithm onBryde’s whales has already been proven in the past, however further developments hadto be made when implementing this algorithm in real-time.The locations of the sensors are determined by fitting each sensor with a GPS module.The data that the GPS module feeds to the system is known as NMEA data, this is usedto track the sensors. The sensor positions as well as detection instances are communicatedto the user-unit for processing via LoRa. LoRa is a useful communication technology asthe data can be transmitted directly over long distances.A time difference of arrival algorithm is run on the user-unit. This algorithm makes useof the data received from the sensors to mathematically locate the source of the audiosignal. The location of the sensors and any potential locations for the signal source isshown graphically via a Python interface on the screen attached to the user unit.After testing it was found that the tracking and communication systems function ade-quately. The DTW based detector proved to run sufficiently, though there were instancesof false-positive detections. Additionally, it was found that the TDOA system tested tobe accurate to within several hundred meters when at a scale of kilometers when testedusing a microcontroller in the laboratory.Through testing the various sub-systems of the overall system appears as it would functionsufficiently, though the system was only laboratory tested and was never deployed in theocean due to time constraints.

AFRIKAANSE OPSOMMING: Die tesis wat aangebied word, gee ‘n indiepte praktiese implementering vir ’n passieweakoestiese deteksietelsel wat gebruik maak van verskil in aankomstyd-teorie om die posisievan klankseine van Bryde se walvisse te lokaliseer. Die stelsel bestaan uit drie individuelesensors wat in die water dryf, wat elk met die gebruikerseenheid op die navorsingsvaar-tuig kommunikeer. Die sensors bestaan uit Raspberry Pis en verskeie doelgerigte kom-ponente.Die oordrag van die sensorlesings en die berekeninge wat met die data gedoenword, maak dit moontlik om die ligging van die klanksein te bepaal deur middel van ’nverskil in aankomstydalgoritme.Die algehele akoestiese lokaliseringstelsel word moontlik gemaak deur verskillende subs-telsels te integreer, naamlik; opsporing, sensorvolging, kommunikasie, tydsinchronisasie,en seinlokalisering. Die teoretiese beginsels onderliggend aan elke substelsel word be-spreek, asook die praktiese implementering, simulasie en toetsing van elke substelsel.Om die oorsprong van ’n klanksein te lokaliseer, moet eers bepaal word of die insetseinna die stelsel die gewenste sein is. ’n Dinamiese tydsverskuiwingsalgoritme is gebruik asdetektor vir die verlangde klankseine. Die geldigheid van hierdie algoritme op Bryde sewalvisse is reeds in die verlede bewys, maar verdere aanpassings is gemaak om hierdiealgoritme intyds te implementeer. Die ligging van die sensors word bepaal deur elke sensor met ‘n GPS toe te rus. Diedata wat die GPS-module na die stelsel voer, staan bekend as NMEA-data. Dit word ge-bruik om die sensors op te spoor. Die sensorposisies sowel as deteksies word deur LORAgekommunikeer na die gebruikerseenheid vir verwerking. LoRa is ’n nuttige kommunikasi-etegnologie wat dit moontlik maak om data direk oor lang afstande oor te dra.’n Verskil in aankomstydalgoritme word op die gebruikerseenheid gebruik. Hierdie algo-ritme maak gebruik van die data wat van die sensors ontvang word om die oorsprongvan die klanksein wiskundig te lokaliseer. Die ligging van die sensors sowel as moontlikeliggings vir die seinbron word grafies vertoon op die skerm van die gebruikerseenheid deurvan ’n Python-koppelvlak gebruik te maak.Na toetsing is bevind dat die opsporing- en kommunikasiestelsels voldoende funksioneer.Die dinamiese tydsverskuiwings-gebaseerde detektor het voldoende gewerk, alhoewel daargevalle was waar vals-positiewe opsporings gemerk is. Daarbenewens is bevind deur lab-oratoriumtoetse met behulp van ‘n mikroverwerker dat die verskil in aankomstydstelselakkuraat is tot binne ’n paar honderd meter, getoets op ’n skaal van kilometers.Deur toetsing van die verskillende substelsels van die algehele stelsel, blyk dit dat diestelsel voldoende funksioneer, alhoewel die stelsel slegs in die laboratorium getoets is ennooit in die oseaan ontplooi is nie as gevolg van tydsbeperkings.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/109783
This item appears in the following collections: