Naïve Bayes For RadarMicro-Doppler Recognition

1. Introduction

Background

Motivation For Work

Review Of Relvant Papers

3. Naïve Bayesian Classifier

Definition

Assumption

Independence

Validity

Improved By PCA

Not Perfect

2. Data

Thales

Micro-Doppler

Classes

Wheeled

Personnel

Tracked

Unknown

Pre-Processing

FFT

Absolute

Decibel

Power

Clutter

Normalizing

Principal Component Analysis

Reference

Test

4. Performance Prediction

Probability Of Error

Bhattacharyya Bound

Cases

Wheeled vs Personnel

Wheeled vs Tracked

Tracked vs Personnel

Vehicles vs Tracked

Limits

Binary

Sub-Optimal

Better

Chernoff

Difficult

Assumptions

Probability Distributions

Multivariate Guassians

Equal Prior Probabilities

Sufficient Data To Estimate μ and Σ

6. Results

Variance

Range

Min 45%

Step 5%

Max 95%

Performance

Parameters

Pcc

Pdec

Pfa

Datasets

Frames

Duration

Reference

Correlation

Size

7. Conclusions

Use Of Bhattacharyya Bound

Performance Of Naïve Bayesian

The Pre Processing Steps

5. Test Strategy

Assessment Of Variance

Analysis Of Classifier at ±5%