A synthesizer framework for multimodal cardiorespiratory signals

Abstract

Multimodal biosignals play an increasing role in medical signal processing. While many novel methods for acquisition and sensorfusion are developed and evaluated, the modeling aspect sees relatively little attention. To overcome this, a synthesizer framework for the generation of multimodal cardiorespiratory signals is presented. The first part of the model consists of a dynamic system of six coupled nonlinear ordinary differential equations. The resulting oscillators generate modality-independent cardiac and respiratory phase signals, which are coupled by respiratory sinus arrhythmia. Moreover, Mayer oscillations of the heart rate are simulated. The second part of the model consists of modality-dependent waveform generators. For each modality, these waveform generators operate with a specific cardiac template that is modulated depending on the respiratory phase signal. The approach is validated qualitatively and quantitatively on three different databases for a variety of biosignals, namely standard electrocardiography (ECG), blood pressure signals, ballistocardiography, photoplethysmography, capacitively coupled ECG as well as respiratory flow and effort. Finally, the possibility to simulate a multimodal recording by combining templates obtained from different databases is demonstrated. The resulting multimodal biosignals mimic important physiological aspects such as modulation due to respiration and exhibit plausible phase relationships. An application of the model to evaluate an algorithm for sensor fusion is demonstrated.