A Fast Analyzing System Fop, Holter Recording Using Digital Signal Processores And Neural Networks
Akira Iwata, T. Sugamoto, Nobuo Suzumura, Yoshitaka Moriguchi · 2005
fast analyzing system for Holter recording using digital signal processors (DSPs) and neural networks has been de- veloped. In the first stage of the analyzing process a 24 hours holter recording on a magnetic tape is replayed in 12 minutes with a high speed tape player. During replaying the detect- ing and matching procedure of QRS-complex waveforms is performed. 4 DSPs are used for computing similarities be- tween electrocardiogram and template waveforms of QRS- complexes. Detected template waveforms, the times and the template numbers when templates match subsequent QRS- complexes are stored in memory. In the second stage the unification process of the detected templates is performed to unify the similar templates. Two neural network models are used to unify the templates. Finally all QRS-complexes are labeled with the indexes of the unified templates, which are utilized for the clinical diagnosis.