Performance analysis of partial self organizing maps for radar signal clustering in electronic warfare
| Autor: | Mustafa Atahan Nuhoglu, Berkin Yildirim |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Self-organizing map
business.industry Computer science 020206 networking & telecommunications Pattern recognition 02 engineering and technology Signal clustering law.invention law Pulse-amplitude modulation Angle of arrival 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Artificial intelligence Electronic warfare Radar business Cluster analysis Block (data storage) |
| Zdroj: | SIU |
| Popis: | Pulse description words, which are information with usage in electronic warfare in order to define radar signals having the following features: pulse width, pulse amplitude, frequency and angle of arrival. Before using this information, radar signals should be deinterleaved by clustering. Performance and running speed of Self Organizing Maps (SOM) algorithm degrade as number of the words and total class number to be clustered in a block increase. In this work, we propose a solution for this problem. Firstly, pulse description words are divided into pieces according to features with the largest variances while total number of neurons stays the same but each part has its own number of neurons and maximum epoch number depending on their data size. Then, each piece is trained by its own and overall neurons from all pieces are connected by defining new neighbourhoods; so that higher performance is achieved in shorter time. The algorithm is called Partial Self Organizing Maps due to the traininig of each piece. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|