# Automatic Bat Species Identification

How automatic bat species identification works. Deep learning models trained on 2.5 million calls achieve 98.9% F1 accuracy. Compare AI classifiers for bat echolocation analysis.

## How It Works

Automatic bat species identification uses machine learning to classify bat echolocation calls to species level. The process: raw ultrasonic recordings are converted to spectrograms, a deep learning model extracts acoustic features (call frequency, duration, shape, inter-pulse interval), and a classifier assigns species probabilities.

## BioSonic's Approach

BioSonic uses convolutional neural networks trained on 2.5 million verified bat calls and 3.5 million noise files. The model identifies species, classifies behaviour (echolocation, feeding buzzes, social calls), and filters non-bat noise with 13.4x fewer false positives than alternatives.

## Accuracy

- 98.9% F1 score in independent benchmarks (Jan Drachmann, Denmark)
- 96.3% balanced accuracy in Swedish Government benchmark (Naturvardsverket)
- 13.4x fewer false positives than Kaleidoscope Pro
- 16.3x fewer missed bats than Kaleidoscope Pro

## Related Pages

- [Benchmarks](https://biosonic.io/benchmarks)
- [Compare Bat Analysis Software](https://biosonic.io/compare)
- [FAQ](https://biosonic.io/faq)