Related Experiment Videos
Robust sound onset detection using leaky integrate-and-fire neurons with depressing synapses
Leslie S Smith1, Dagmar S Fraser
1Department of Computing Science and Mathematics, University of Stirling, Stirling FK9 4LA, UK. lss@cs.stir.ac.uk
IEEE Transactions on Neural Networks
|October 16, 2004
Summary
This study introduces a biologically inspired method for detecting sound onsets using auditory nerve-like spikes and a specialized neuron. The technique achieves near-instantaneous onset detection across various sound types.
Area of Science:
- Neuroscience
- Signal Processing
- Bioacoustics
Background:
- Accurate sound onset detection is crucial for auditory perception.
- Existing methods may have latency issues.
- Biological auditory systems offer efficient models for signal processing.
Purpose of the Study:
- To develop a biologically inspired computational model for rapid sound onset detection.
- To simulate auditory nerve (AN) spike coding and synaptic transmission.
- To evaluate the model's performance on diverse auditory stimuli.
Main Methods:
- Utilized a cochlea-like filter to process input sounds.
- Employed spike coding to mimic auditory nerve activity.
- Simulated a leaky integrate-and-fire neuron with a depressing synapse.
- Tested the model with tone bursts, musical sounds, and the TIMIT speech corpus.
Main Results:
- Achieved essentially zero-latency onset detection relative to AN-like spikes.
- Demonstrated effective onset detection across different sound categories.
- Validated the biological plausibility of the proposed neural network architecture.
Conclusions:
- The biologically inspired approach offers a highly efficient and low-latency solution for sound onset detection.
- The model provides insights into neural mechanisms of auditory signal processing.
- This technique has potential applications in hearing prosthetics and speech recognition systems.