Scalable multidimensional relations between features of calls to emergency services

Quite frequent calls to emergency telephone services with false messages form a real problem. They mainly come from unreliable speakers or are subject of multiple mutually inconsistent reports of the same event. The paper presents an automated support for recognition of such calls. In order to reflect real cases precisely, we propose and formulate multidimensional call features correlations. This approach significantly improves the quality of results, proved by experiments. The more features are taken into comparison, the better the results reflect the real cases. With the proposed approach, it is possible to scale the accuracy of recognition. The phone operator, adequately to a defined reliability of the input data, can choose a properly needed level of accuracy. The proposed analysis can be applied not only in the emergency services, but also in many other fields of interest, in which multidimensional data should be compared.