Performance evaluation of BP and PCA neural networks for ECG data compression
Yasunori Nagasaka, Akira Iwata · 2005
The performances of a BPNN (neural network trained by back-propagation) and a PCANN (neural network which computes principal component analysis) in the authors' ECG data compression system have been investigated by comparing them with an existing method, TOMEK. The authors used the MIT/BIH Arrythmia database as ECG data. Both BPNN and PCANN showed 1.1 to 1.4 times higher compression rates than TOMEK with 13.0% of PRD and 99.0% of CC. BPNN was a little better than PCANN in compression rate. Observing the reproduced waveforms, BPNN and PCANN had almost the same performance, and they were superior to TOMEK. Although the difference is small, the performance of BPNN was always superior to PCANN.