Liquid State Machine for Speech Classification

This repository contains the MATLAB implementation of a Liquid State Machine (LSM) for speech classification in Triton folder, demonstrating the use of an approximate state-space model for predictive performance, as described in the papers listed in the citation section.

Citing

If you find this repository useful in your research, please consider citing the following papers:

@inproceedings{gorad2019predicting,
  title={Predicting performance using approximate state space model for liquid state machines},
  author={Gorad, Ajinkya and Saraswat, Vivek and Ganguly, Udayan},
  booktitle={2019 International Joint Conference on Neural Networks (IJCNN)},
  pages={1--8},
  year={2019},
  organization={IEEE}
}

@inproceedings{saraswat2021hardware,
  title={Hardware-friendly synaptic orders and timescales in liquid state machines for speech classification},
  author={Saraswat, Vivek and Gorad, Ajinkya and Naik, Anand and Patil, Aakash and Ganguly, Udayan},
  booktitle={2021 International Joint Conference on Neural Networks (IJCNN)},
  pages={1--8},
  year={2021},
  organization={IEEE}
}

Usage

Set the simulation parameters in ExamplePrescript.m. Run the main.m script to start the LSM simulation, which in turn calls SpokenDigitsLSM.m.

Prerequisites

MATLAB R2021a or later Auditory Toolbox for MATLAB (included in the repository)

Installation

Clone the repository and run the main.m script in MATLAB.

Output

The simulation outputs results in a .mat file along with log files.