Related Experiment Videos
Multistage classification and recognition that employs vector quantization coding and criteria extracted from
Manal M Abdelwahab1, Wasfy B Mikhael
1School of Electrical Engineering and Computer Science, University of Central Florida, Orlando, Florida 32816-2450, USA.
Applied Optics
|January 23, 2004
Summary
This study introduces a self-designing system using classification decision trees for effective signal recognition. The novel approach enhances accuracy, even with noisy data, by optimizing criteria and using vector quantization.
Area of Science:
- Computer Science
- Signal Processing
- Machine Learning
Background:
- Classification decision tree algorithms are increasingly applied to pattern recognition.
- Effective recognition of a large number of signals remains a challenge in signal processing.
Purpose of the Study:
- To propose a self-designing system for signal recognition using classification decision trees.
- To enhance signal recognition effectiveness through preprocessing and optimized criteria selection.
Main Methods:
- Utilizing classification tree algorithms for signal recognition.
- Applying preprocessing techniques to original and transformed signal domains.
- Optimizing criteria at each tree node for complexity and noise immunity.
- Employing vector quantization to group signals at each stage.
Main Results:
- The system achieves excellent classification accuracy for a large number of signals.
- High performance is maintained even with noisy and corrupt data.
- Each signal is uniquely represented by a composite binary word index.
Conclusions:
- The proposed self-designing system demonstrates robust and accurate signal recognition capabilities.
- The method is effective in handling challenging conditions, including noisy and corrupted signals.