Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Acoustic Identification of Ae. aegypti Mosquitoes using Smartphone Apps and Residual Convolutional Neural Networks

Kayuã Oleques Paim, Ricardo Rohweder, Mariana Recamonde-Mendoza, Rodrigo Brandão Mansilha1, Weverton Cordeiro

TL;DR

The paper tackles low-cost, smartphone-based acoustic identification of Ae. aegypti by analyzing wingbeat sounds with CNNs. It develops three classifier paradigms—binary, multiclass, and an ensemble of binary models—evaluated on a wingbeat audio dataset derived from the Abuzz project, with careful pre-processing and feature extraction to mel-scaled spectrograms. The results show high binary accuracy (~97.7%) and strong ensemble recall (~96.8%), while the multiclass approach provides useful species-level insights but with lower overall accuracy (~78%). The work demonstrates the potential of accessible, crowd-sourced surveillance tools for mapping mosquito incidence and guiding public health actions, while outlining practical challenges for offline deployment and UI design.

Abstract

In this paper, we advocate in favor of smartphone apps as low-cost, easy-to-deploy solution for raising awareness among the population on the proliferation of Aedes aegypti mosquitoes. Nevertheless, devising such a smartphone app is challenging, for many reasons, including the required maturity level of techniques for identifying mosquitoes based on features that can be captured using smartphone resources. In this paper, we identify a set of (non-exhaustive) requirements that smartphone apps must meet to become an effective tooling in the fight against Ae. aegypti, and advance the state-of-the-art with (i) a residual convolutional neural network for classifying Ae. aegypti mosquitoes from their wingbeat sound, (ii) a methodology for reducing the influence of background noise in the classification process, and (iii) a dataset for benchmarking solutions for detecting Ae. aegypti mosquitoes from wingbeat sound recordings. From the analysis of accuracy and recall, we provide evidence that convolutional neural networks have potential as a cornerstone for tracking mosquito apps for smartphones.

Acoustic Identification of Ae. aegypti Mosquitoes using Smartphone Apps and Residual Convolutional Neural Networks

TL;DR

The paper tackles low-cost, smartphone-based acoustic identification of Ae. aegypti by analyzing wingbeat sounds with CNNs. It develops three classifier paradigms—binary, multiclass, and an ensemble of binary models—evaluated on a wingbeat audio dataset derived from the Abuzz project, with careful pre-processing and feature extraction to mel-scaled spectrograms. The results show high binary accuracy (~97.7%) and strong ensemble recall (~96.8%), while the multiclass approach provides useful species-level insights but with lower overall accuracy (~78%). The work demonstrates the potential of accessible, crowd-sourced surveillance tools for mapping mosquito incidence and guiding public health actions, while outlining practical challenges for offline deployment and UI design.

Abstract

In this paper, we advocate in favor of smartphone apps as low-cost, easy-to-deploy solution for raising awareness among the population on the proliferation of Aedes aegypti mosquitoes. Nevertheless, devising such a smartphone app is challenging, for many reasons, including the required maturity level of techniques for identifying mosquitoes based on features that can be captured using smartphone resources. In this paper, we identify a set of (non-exhaustive) requirements that smartphone apps must meet to become an effective tooling in the fight against Ae. aegypti, and advance the state-of-the-art with (i) a residual convolutional neural network for classifying Ae. aegypti mosquitoes from their wingbeat sound, (ii) a methodology for reducing the influence of background noise in the classification process, and (iii) a dataset for benchmarking solutions for detecting Ae. aegypti mosquitoes from wingbeat sound recordings. From the analysis of accuracy and recall, we provide evidence that convolutional neural networks have potential as a cornerstone for tracking mosquito apps for smartphones.
Paper Structure (17 sections, 2 equations, 10 figures, 4 tables)

This paper contains 17 sections, 2 equations, 10 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Materials and Methods
  4. Data Collection and Pre-processing
  5. Features extraction
  6. Convolutional Neural Networks
  7. Model Evaluation
  8. Proposed Approach
  9. Binary Classifier
  10. Multiclass Classifier
  11. Ensemble of Binary Classifiers
  12. Experiments and Results
  13. Analysis of Fourier Transform Parameters
  14. Performance analysis of the binary classifier
  15. Performance analysis of the multiclass classifier
  16. ...and 2 more sections

Figures (10)

  • Figure 1: Distribution of audio duration and frequency in the original dataset with mosquito wingbeat sounds.
  • Figure 2: Workflow considered in the analysis of the proposed methodology.
  • Figure 3: Performance distribution of accuracy, precision, recall, and F1-measure for the model configurations tested (Table \ref{['tab:eval:list-configurations']}). In summary, config. #8 presented the best overall performance and was selected for further experiments.
  • Figure 4: Average and standard deviation for performance metrics analyzed for each model configuration.
  • Figure 5: Performance distribution for the binary model using FFT parameters configuration #8. The black diamonds indicate the mean for each performance metric.
  • ...and 5 more figures