Effects of interpulse-interval variation on deep-learning classification of bat vocalizations

2026-10-01

AI Summary

This study investigates how inter-pulse interval (IPI) variability affects the performance of deep learning models in classifying bat echolocation calls. To this end, natural and normalized IPI datasets were constructed, and controlled experiments were conducted using EfficientNet-B0 and PaSST architectures to systematically quantify the differential sensitivities of convolutional neural networks and Transformers to temporal context. The findings reveal that natural IPIs contribute minimally to species classification; however, models trained on normalized data exhibit a significant accuracy degradation when evaluated on natural recordings. These results underscore the critical importance of distributional alignment between training and testing data, offering a novel perspective for evaluating the cross-condition generalization capabilities of bioacoustic models.

Abstract

Temporal context may aid automated bat-species classification, but the contribution of specific features remains unclear. We investigated whether variation in the interpulse interval (IPI)-the time between consecutive call onsets-provides species-discriminative information and whether transformer-based models are more sensitive to this information than convolutional neural networks. We created two matched datasets from European bat recordings: a natural-IPI condition retaining the original call timing and a normalized-IPI condition in which call onsets were spaced at 50-ms intervals. EfficientNet-B0 and PaSST were fine-tuned and evaluated within each condition. In an additional experiment, each architecture was trained separately on natural-IPI and normalized-IPI recordings, and evaluated on the same natural-IPI test set. Finally, the pretrained classifiers BatDetect2 and BAT were evaluated on both conditions. Within-condition IPI normalization had model-dependent effects. PaSST accuracy differed little between the natural-IPI ($71 \pm 2.3\%$) and normalized-IPI ($70 \pm 6.3\%$) conditions, whereas EfficientNet accuracy increased from $47 \pm 4.7\%$ to $57 \pm 3.9\%$. PaSST exceeded EfficientNet under both conditions. In the cross-condition evaluation, models trained on natural-IPI recordings outperformed those trained on normalized-IPI recordings on the natural-IPI test set: accuracy decreased from 54% to 50% for EfficientNet and from 65% to 57% for PaSST. BatDetect2 and BAT differed little between IPI conditions. Overall, we found limited support for the hypotheses that natural IPI variation contributes substantially to bat-species classification and that it is used more effectively by transformer-based than CNN-based models. Nevertheless, the cross-condition performance decrease shows that results obtained under normalized conditions may not transfer fully to natural recordings.

Problem

• bat vocalization classification
• interpulse interval
• deep learning
• species discrimination
• temporal context

Innovation

• Interpulse interval
• Transformer
• Convolutional neural network
• Bioacoustics classification
• Cross-condition evaluation