Library sequential search

A normalized scalar product of two spectral vectors was used during matching as the similarity measure, and sequential searches through the entire library were always performed. Thus, each spectrum in turn was treated as a query. To simulate small variances in data acquisition, and/or spectral differences for very similar compounds, 1% and 5% of random white noise, has been added. Appropriate decomposition (wavelet or PCA) was then performed, and the resulting vector was compared to each of the spectral vectors in the compressed library. 

This kind of search is based on comparing the measured spectrum with candidate spectra of the library, bit by bit. Sequential search is only useful if a small data set is to be treated or if it is obUgatory to retrieve every individual data set. A more efficient way is to sort the entities in a database by deriving appropriate keys. 

In addition to the necessity of defining similarity, one other major choice must be made in developing library searching approaches. This involves the use of sequential search, the use of inverted (sorted) files, employing hashing methods, or the use of hierarchical trees. Each of these methods has its advantages and is suitable in some circumstances. 

A sequential search of a spectral library, for example, involves the comparison of every part of a sample spectrum with library spectra, and is suitable only for small libraries. A hierarchical search involves comparing groups (families) of spectra having the same set of key features as the sample spectrum, enabling large libraries classified in a tree-like structure to be searched very efficiently. 

Table 5.9 summarises the main features of FTIR spectroscopy as applied to extracts (separated or not). Since many additives have quite different absorbance profiles FTIR is an excellent tool for recognition. Qualitative identification is relatively straightforward for the different classes of additives. Library searching entails a sequential, point-by-point, statistical correlation analysis of the unknown spectrum with each of the spectra in the library. Fully automated analysis of... 

When peaks are incompletely separated identification may still be possible using a reverse search. The ability of an algorithm to match two or more components in the mass spectrum of a mixture is aided by requiring only that the peaks of the reference spectrum are present in the unknown spectrum rather than the other way round, as for a normal (or forward) search. The hit list of retrieved library spectra should then represent the compounds in the spectrum of the mixture provided that their spectra are present in the reference library. Subtracting the best-hit library spectrum from the mixture spectrum produces a residual spectrum that can then be matched against the other spectra in the hit list in a forward search. In a sequential process identification of the component spectra may be achieved. 

Computer library searching of spectral databases is routinely available. The database is usually a component part of the spectrometer although the search may be undertaken remotely. Several attempts have been made to develop artificial intelligence systems for direct spectral interpretation, but to date these have met with limited success. Advances in computer control have allowed multiexperiment analysis in which the spectrometer will follow a set of experiments sequentially while automatically adjusting operating parameters as directed by the results of the preceding experiment. Further advances in this area are anticipated. 

---