Started Working as a Research Assistant at the Data Analytics and Visualization (DAV) Lab at UCSC

Tags: UCSC, Research

Research Proposal for DAV Lab

Title: Towards Biodiversity Monitoring through Bird Species Identification by Bird Sounds Classification

Background

Bird diversity is a well-established indicator of ecosystem health due to their varied habitat preferences and high mobility. Traditional observer-based surveys, while valuable, are often limited in scope by logistical and financial constraints, especially for large-scale biodiversity monitoring.

Passive acoustic monitoring (PAM) offers a promising alternative. By combining soundscape recordings with advanced machine-learning techniques, researchers can achieve greater spatial coverage and temporal resolution in biodiversity assessments. This allows for a more in-depth exploration of the relationship between restoration efforts and biodiversity changes. In the face of escalating anthropogenic pressures, such as habitat modification and climate change, employing cutting-edge tools for biodiversity monitoring is crucial. This research builds upon the inspiration provided by the ongoing BirdCLEF competition, a Kaggle competition focused on automated bird species identification from soundscapes in the Western Ghats of India.

The BirdCLEF competition highlights the potential of PAM for biodiversity monitoring, particularly its ability to address challenges like identifying endemic species, detecting data-deficient endangered birds, and classifying poorly understood nocturnal birds. This research aims to develop a similar framework for Sri Lanka, utilizing PAM and machine learning to establish an automated system for bird species detection, classification, and monitoring. By doing so, this research will contribute to improved biodiversity monitoring and conservation efforts in Sri Lanka and also this research will contribute to the tourism industry in Sri Lanka by promoting and facilitating ecotourism.

Research Aim

Detects, identifies and monitors birds in Sri Lanka by continuously recording natural soundscapes over long periods.

Research Approach:

This will be a set of research projects that includes data mining, data analytics, machine learning, digital signal processing, etc. The following proposed research questions can be addressed by experimental research with cross-sectional quantitative data.

Research Questions:

  1. What are the unique audio signal features (time domain and frequency domain) of the vocalization of different bird species?
  2. How can different bird species be accurately identified by using a long audio file effectively?
  3. What are the noise removal techniques to extract bird vocalization from a given audio clip?
  4. What is the best DNN architecture to develop a model to classify birds by their vocalizations?
  5. How can audio signal analysis be effectively used to monitor bird species diversity?
  6. How can audio signal analysis be effectively used to monitor bird activities (finding food, making nests, defending, attacking, etc.)

Research Outcomes:

  1. A scientific system to identify bird species by bird vocalization.
  2. A scientific system to identify the diversity of bird vocalization within the same bird species.
  3. A data mining platform to identify areas and time periods rich in different bird species.
  4. A system to effectively monitor bird diversity and bird activities in different locations in Sri Lanka.
  5. A mobile app to spot and track different bird species by their sound.

Plans for Sustainability:

  • This research can lead to advancements in the field of DSP and ML. The development of accurate and efficient algorithms for bird species identification can have broader applications in other fields such as speech recognition, music analysis, etc.
  • This research can contribute significantly to biodiversity conservation efforts. Changes in bird populations can provide early warnings about the state of our ecosystems. The data generated from these research outcomes can be promoted for making national policies and conservation strategies. It can help in prioritizing areas for conservation, developing effective management plans, and evaluating the impact of conservation interventions.
  • The integration of ecological knowledge with advanced technological tools can open new avenues in environmental research. It can lead to more comprehensive and accurate studies on various aspects of ecology and biodiversity.
  • The research outcomes will be used to digitally identify newly migrated birds and bird migration patterns. Therefore, this system can be applicable for worldwide bird monitoring tasks.

Plan for Commercialization:

  • The data generated from this research can be used to identify areas rich in bird diversity with time periods in the year, which can be promoted as bird-watching spots and seasons for eco-tourists.

  • The bird species identification system can be integrated into a mobile application. When visitors are in nature parks or reserves, the app can notify them when a recognized bird species is nearby, enhancing their visitor experience.

Timeline:

References:

Birdclef (2024) Kaggle. Available at: https://www.kaggle.com/competitions/birdclef-2024/overview (Accessed: 14 May 2024).

Bird-vocalization-classifier (2023.02.04) Kaggle. Available at: https://www.kaggle.com/models/google/bird-vocalization-classifier/TensorFlow2/bird-vocalization-classifier (Accessed: 14 May 2024).

Back to top