Adaptive representations of sound for automatic insect recognition

Abstract

Insect population numbers and biodiversity have been rapidly declining with time, and monitoring these trends has become increasingly important for conservation measures to be effectively implemented. But monitoring methods are often invasive, time and resource intense, and prone to various biases. Many insect species produce characteristic sounds that can easily be detected and recorded without large cost or effort. Using deep learning methods, insect sounds from field recordings could be automatically detected and classified to monitor biodiversity and species distribution ranges. We implement this using recently published datasets of insect sounds (up to 66 species of Orthoptera and Cicadidae) and machine learning methods and evaluate their potential for acoustic insect monitoring. We compare the performance of the conventional spectrogram-based audio representation against LEAF, a new adaptive and waveform-based frontend. LEAF achieved better classification performance than the mel-spectrogram frontend by adapting its feature extraction parameters during training. This result is encouraging for future implementations of deep learning technology for automatic insect sound recognition, especially as larger datasets become available.

Extensive data engineering to the rescue: building a multi-species katydid detector from unbalanced, atypical training datasets

“…A meaningful expansion of the playback set with additional recaptured recordings may help stabilize performance improvements. Alternatively, one could also simulate a playback set by convolving the focal recordings with impulse responses representative of the target environments that one attempts to mimic [ 72 ]. Such a technique offers a cost-effective alternative to using physical transmitters and receivers and offers the ability to ‘add’ newer environments to a playback set without the need for physical access to those environments.…”

Towards a toolkit for global insect biodiversity monitoring

“…Passive acoustic monitoring employs autonomous recording units (ARUs), which record the sounds of soniferous insects and other animals either continuously or intermittently (subsamples of minutes) in both the audible and ultrasonic spectrum [ 44 ]. Recorded sounds are usually automatically identified using artificial intelligence (AI) trained on annotated sounds from known species [ 45 ]. Newer and cheaper ARUs and better analytical methods, such as deep learning, have allowed for acoustic monitoring at unprecedented temporal and spatial scales [ 17 ], as well as the re-analysis of previously recorded sounds (i.e.…”

Acoustic monitoring for tropical insect conservation

“…Sugai et al 2019, Gibb et al 2019) in combination with curated multimedia repositories with references to voucher specimens, as well as pictures and videos provided by citizen scientists. In addition, considerable progress has been made in automatic song recognition, which allows identification of prominent songsters in large datasets from an increasing number of acoustic monitoring sites (Faiß and Stowell 2023; Madhusudhana et al 2024).…”

“…Much of the focus of automated recognition has been on vertebrate taxa, especially birds and mammals. Recently however, automatic recognition algorithms have been developed to discriminate different mosquito and bee species from their flight sounds (Kawakita and Ichikawa 2019), and cicada, cricket and katydid species based on their calls (Noda et al 2019, Tey et al 2022, Faiß and Stowell 2023), some with impressive accuracies of species-level discrimination (90-98%). Most of the algorithms have however been developed using recordings of individual species available in databases, with data augmentation techniques to introduce noise.…”