Data compression for audio-based smart beekeeping
Loading...
Supplementary material
Other Title
Authors
Shi, Guangyu
Author ORCID Profiles (clickable)
Degree
Master of Applied Technologies (Computing)
Grantor
Unitec, Te Pūkenga – New Zealand Institute of Skills and Technology
Date
2024
Supervisors
Song, Lei
Ardekani, Iman
Ardekani, Iman
Type
Masters Thesis
Ngā Upoko Tukutuku (Māori subject headings)
Keyword
New Zealand
beehive monitoring
honeybees (Apis mellifera)
MP3 (Audio coding standard)
FLAC (Audio coding standard)
data compression (Computer science)
audio data processing systems
apiculture industry
AI in agriculture
beehive monitoring
honeybees (Apis mellifera)
MP3 (Audio coding standard)
FLAC (Audio coding standard)
data compression (Computer science)
audio data processing systems
apiculture industry
AI in agriculture
ANZSRC Field of Research Code (2020)
Citation
Shi, G. (2024). Data compression for audio-based smart beekeeping (Unpublished document submitted in partial fulfilment of the requirements for the degree of Master of Applied Technologies (Computing)). Unitec, Te Pūkenga - New Zealand Institute of Skills and Technology https://hdl.handle.net/10652/6491
Abstract
RESEARCH QUESTIONS
• Question 1
How does FLAC compression impact the classification accuracy of different bee activities compared to uncompressed and MP3 compressed formats?
• Question 2
What are the storage and transmission efficiency gains achieved by using FLAC in beehive monitoring?
ABSTRACT
The research aimed to address the challenges associated with audio data compression in beehive monitoring by exploring the feasibility and effectiveness of using the Free Lossless Audio Codec compression format. The contribution is demonstrated by the efficacy of FLAC compression in reducing resource consumption without feature loss, thereby not compromising AI performance. The methodology involved using FLAC techniques for audio compression, extracting relevant acoustic features using Mel-Frequency Cepstral Coefficients, and implementing Support Vector Machine models to classify and analyse hive conditions. The results demonstrated that Free Lossless Audio Codec outperformed MPEG-1 Audio Layer 3 and uncompressed Waveform Audio File formats in maintaining the efficiency of audio signals and the integrity of critical acoustic features. These metrics include waveform characteristics, classifier accuracy, compression degree and speed, and transmission speed through the inclusion of multiple data sources. The findings highlight Free Lossless Audio Codec as an advantageous option for beehive monitoring systems. Despite the positive findings, the research has some limitations. The datasets used in the experiments may not encompass all possible beehive conditions and environmental variations. Additionally, the SVM models were implemented with specific parameters that may not generalize to all contexts. Further research is needed to assess the robustness of the findings across a broader range of conditions and to explore different machine learning approaches.
Publisher
Permanent link
Link to ePress publication
DOI
Copyright holder
Author
Copyright notice
All rights reserved