Olfactory target/background odor detection via self-expression model

Extreme Learning Machine (ELM) is a promising single hidden layer feed-forward neural network learning method, which achieves fast learning by randomly tuning the hidden layer. In this paper, we propose a self-expression ELM (SeELM) for olfactory target/background detection. Specially, it is known that metal oxide semiconductor (MOS) sensor in electronic nose (E-nose) can response to several target gases and interferences (background) simultaneously, which would seriously deteriorate the detection accuracy of target gases. Considering that there are numerous interferences in real-world application scenario, it is impossible for us to collect them. With a prior knowledge that the target gases samples can be easily collected, a novel SeELM method is proposed to address this issue. The idea is represented as two aspects. First, the target gases being detected by an E-nose can be fixed as invariant information, which is utilized to construct a self-expression model. Second, with the self-expression detector, the unknown backgrounds can be easily recognized in terms of the violation. Experimental results proved that the proposed SeELM method is significantly effective for Target/Background detection in E-nose.