A multisensor data fusion approach for improving the classification accuracy of uterine EMG signals

Multisensor data fusion is an important technique used for solving various pattern recognition problems. In this paper, we used data fusion for improving the classification of uterine electromyogram (EMG) signals recorded by 16 electrodes positioned on the abdominal wall of the pregnant women. First, we evaluated the classification performance of each channel. Then, we applied a decision-level fusion method based first on the majority voting (MV), then on the weighted majority voting (WMV) rules. The results were very promising. The fusion of data from multiple sensors improved the accuracy of uterine EMG classification. The high percentage of correctly classified events, compared with earlier results, proves the efficiency of this approach for detecting labor.