All Projects

SNN / DVS Projects

Spoken Digits Live SNN Inference

A real-time inference pipeline that classifies spoken digits using spiking neural representations and temporal feature processing — running live from a microphone input.

SNNAudioLive Inference

Live Demo / Screen Recording placeholder

Replace with an embedded video or interactive inference demo.

Overview

This project implements a complete, end-to-end pipeline for real-time spoken digit recognition using spiking neural networks. Audio captured from a microphone is converted into spike trains via a cochlear-inspired encoding stage, then fed into a trained SNN that produces a classification within a single inference window.

Pipeline

  1. 1Microphone capture and buffering at configurable sample rates.
  2. 2Cochlear filterbank encoding — audio to spike trains.
  3. 3Feed-forward SNN inference over a fixed time window.
  4. 4Softmax readout and digit label display in real time.

Network Architecture

Placeholder — describe the SNN topology, neuron model (LIF / adaptive LIF), number of layers, time steps, and training procedure (e.g., STBP, surrogate gradient).

Results

Placeholder for accuracy, latency, and spike efficiency metrics. Add a confusion matrix or accuracy-vs-timestep plot here.