| abstract |
A method and a system using that method are provided which employ a pattern recognition algorithm to improve sensitivity in detecting hazardous vapors. The algorithm enables the discrimination of vapors of interest from non-hazardous substances at higher concentrations in varying relative humidity. A weight vector is generated corresponding to a N-space representation of a class comprising known vapors of interest, and a N-space representation of the unknown vapor is used to generate an unknown pattern vector. By calculating the dot product of the unknown pattern vector and the weight vector a determination can be made as to whether the unknown vapor is within the class. The weight vector is generated by selecting a training set comprising a subset of the known vapors of interest and background vapors and generating an N-space representation of the training set so as to create an associated weight vector. |