Pattern Recognition Toolbox

PRT is a generalized Pattern Recognition Toolbox for analysis and fault diagnosis of one-dimensional signals using MATLAB as its computing engine. The toolbox was written by engineers versed in applying pattern recognition and neural networks to a wide range of research and product development applications.

An extensive library of feature extraction and optimization routines, statistical classifications algorithms as well as neural networks, and graphical display functions for visualizing data sets are provided. These tools help researchers in determining key relationships and the statistical nature of the data. A Matlab and Java User Interface is provided for automatic data acquisition and analysis. The user-interface supports specific types of data acquisition hardware and I/O file formats. Customized analysis routines are available!

Feature Analysis Tools

• Feature Optimization
• Principal Component Analysis
• Histogram Visualization
• Correlation Visualization

Classical Methods

• K-Nearest Neighbor
• Centroid
• K-Means Clustering
• Threshold Algorithm

Neural Networks

• Backpropagation Network
• Radial Basis Function
• Probabilistic Neural Network

• Frequency Features
• Amplitude Ratio
• Absolute Energy
• % Energy
• Energy Difference
• Energy Ratio
• Spectral Correlation

Statistical Features

• Mean
• Standard Deviation
• Skewness
• Kurtosis

• Time to Peak Value
• Local Damping