Mosquito Detection with Neural Networks: The Buzz of Deep Learning

Many real-world time-series analysis problems are charac- terised by scarce data. Solutions typically rely on hand-crafted features extracted from the time or frequency domain allied with classification or regression engines which condition on this (often low-dimensional) fea- ture vector. The huge advances enjoyed by many application domains in recent years have been fuelled by the use of deep learning architectures trained on large data sets. This paper presents an application of deep learning for acoustic event detection in a challenging, data-scarce, real- world problem. Our candidate challenge is to accurately detect the pres- ence of a mosquito from its acoustic signature. We develop convolutional neural networks (CNNs) operating on wavelet transformations of audio recordings. Furthermore, we interrogate the network’s predictive power by visualising statistics of network-excitatory samples. These visualisa- tions offer a deep insight into the relative informativeness of components in the detection problem. We include comparisons with conventional clas- sifiers, conditioned on both hand-tuned and generic features, to stress the strength of automatic deep feature learning. Detection is achieved with performance metrics significantly surpassing those of existing al- gorithmic methods, as well as marginally exceeding those attained by individual human experts. The data and software related to this paper are available at http://humbug.ac.uk/kiskin2017/.