Stellar data classification using SVM with wavelet transformation
Ping Guo, Fei Xing, Yu–Gang Jiang · 2005
This paper presents a novel stellar spectra recognition technique, which is based on a wavelet transform and support vector machines. Due to the very low signal-to-noise ratio of real world spectral data, a de-noising method for stellar spectra is proposed using a wavelet transform based on the traditional threshold technique. Then support vector machines are adopted to complete the classification. Features in the spatial and wavelet domain are extracted and then used as input of support vector machines. Experimental results show that our technique is robust against noise and efficient in computation. The obtained correct classification rate of the proposed methods is much higher than using either a support vector machine alone or the principle component analysis feature extraction method.