L.A. Manilo - Dr.Sc. (Eng.), Professor, Department of Biotechnical Systems, Saint Petersburg Electrotechnical University «LETI» E-mail: lmanilo@yandex.ru

The problem of linear decision functions creation for several classes of biomedical signals recognition is considered. Importance of the solution for the problem of dangerous arrhythmias detection at early stages of their occurrence is shown. The conclusion about the expediency of the electrocardiogram analysis in frequency domain with application of the multiple discriminant analysis is drawn. For the purpose of approach of Fischer-s criterion to the groups of signals classification accuracy it is offered to use weight functions. For the problem of electrocardiogram recognition using the parameters of a normalized spectrum the formula of weight coefficients calculation is received. The special attention is paid to the procedure of spectral feature space transformation which allows to reduce its dimension and to simplify the algorithm of decision functions creation. Results of the experimental research on different types of dangerous arrhythmias recognition are given. In particular the problems of ventricular fibrillation detection, and also the recognition of arrhythmias which are harbingers of dangerous violations by electrocardiogram are solved. It is shown that use of the Fischer-s discriminant taking into account the weight functions allows to reduce classification errors. The conclusion about high reliability of ventricular fibrillation detection according to the electrocardiogram spectral description is drawn.

 

1. Clayton R.H., Murray A., Campbell R.W.F. Frequency Ana­lysis of Self-Terminating Ventricular Fibrillation // Computers in Cardiology. IEEE Computer Society Press. 1995. P. 705-708.
2. Clayton R.H., Campbell R.W.F., Murray A. Time-Frequency Analysis of Human Polymorphic Ventricular Tachycardia // Computers in Cardiology. IEEE Computer Society Press. 1997. V. 24. P. 97-100.
3. Duda Z., KHart P. Raspoznavanie obrazov i analiz scen / Per. s angl. M.: Mir. 1976. 511 s.
4. Loog M., Duin R.P.W., Haeb-Umbach R. Multiclass Linear Dimension Reduction by Weighted Pairwise Fisher Criteria // IEEE Transactions on Pattern Analysis and Machine Intelligence. 2001. V. 23. № 7. P. 762-766.
5. Manilo L.A. Uporjadochenie spektralnykh priznakov po ehmpiricheskim ocenkam mezhgruppovogo rasstojanija v zadachakh klassifikacii biosignalov // Izvestija vuzov Rossii. Radioehlektronika. 2006. Vyp. 3. S. 20-29.