Data Fusion Design Service from Cambridge Algorithmica
Cambridge Algorithmica offers a specialist service is for design of high-performance Multi-Feature and other Data Fusion. This service is suitable for the following sorts of organisations, as a cost-efficient approach:
The Company effectively uses an alternative to other multi-dimensional statistical pattern matching. It makes many less assumptions than most alternatives, on feature PDFs and the correlations between them. It also treats raw features, pre-processed features and fused features in a unified way. Other approaches can also be incorporated as pre-processing stages. It uses the same underlying powerful techniques proposed initially for multi-modal biometric, under the name Biometric Gain against Impostors (BGI).
The basic approach is firmly grounded on Bayesian statistical techniques, in which the fusion method is specifically tuned to data features in training samples. This ensures robust and straightforward fusion in most circumstances; it also avoids the worse excesses of the curse of dimensionality. Additional techniques allow automatic handling of missing data, whether this is in the training data or the classification data. There is also a combined approach to the handling of noisy features and to the handling of specific noise estimates for all or for particular features.
Finally, Cambridge Algorithmica uses its in-house technical expertise to design fusion that avoids many of the pitfalls of over-simplistic automated processing; this is especially beneficial for the more complicated cases.
The design service is suitable for all the following uses:
Multi-Feature Data Fusion
Multi-Modal Data Fusion
Multi-Instance Data Fusion
Multi-Algorithmic Data Fusion
Multi-Sensorial Data Fusion
Cambridge Algorithmica's service is provided on a bespoke basis, to suit the particular needs of their clients. For further information, please contact Nigel Sedgwick.