|
Full Spectrum Matching Method
This algorithm is very similar to the Euclidean Distance algorithm with one very important difference; both the unknown and the library data are centered about their respective means before the vector dot products are calculated:
where the vectors are defined as:
Where Lib is the library entry being searched and Unkn is the unknown spectrum. In the case where the signal to noise ratio and baselines in the library and unknown spectra is very similar, the mean values will be effectively the same. Note that if the mean centering step is removed, this equation reduces to the square of the Euclidean Distance. However, the mean centering step has one very important effect; the Hit Quality Index is independent of the normalization of the spectra. In Euclidean Distance, the spectra must be normalized between 0 and 1 before the calculation. If the spectrum is noisy, or there are sharp negative dips in the spectrum (i.e., water vapor, CO2, etc.) they will be normalized to 0. This causes the baseline in the unknown to be normalized to a value significantly greater than 0. Therefore, the library spectrum and unknown spectrum will not match very well, and incorrect hits will be given. By centering each spectrum about its mean, this problem is avoided.
In almost all cases, Correlation searching will provide better or equal hit quality information when compared to Euclidean Distance searching. However, due to the extra mean centering step, expect the Correlation search to be slightly slower. The Mask function is applied to both the unknown and the library before the mean centering is calculated to ensure that the masked points don't contribute to the final HQI. | 
0 Items