Respiratory impedance spectral estimation for digitally created random noise.

摘要: Measurement of respiratory input mechanical impedance (Zrs) is noninvasive, requires minimal subject cooperation, and contains information related to lung function. A common approach measure Zrs apply random noise pressure signals at the airway opening, resulting flow variations, then estimate using Fast-Fourier Transform (FFT) techniques. The goal this study was quantify how several signal processing issues affect quality a spectral when sequence created digitally. Random driven time domain data were simulated for three models, which permitted predictions characteristics previously reported from 0–4, 4–32, 4–200 Hz. Then, evaluated as function number runs ensemble averaged, windowing, signal-to-noise ratio (SNR), magnitude shape |P(jω)|. For |P(jω)| with uniform power distribution SNR 200 10 minimum γ2 >0.95. However, matched |Zrs|, > 0.91 even 5 20. digitally equally spaced content, rectangular window superior Hanning. Finally, coherence alone may not be reliable because only an itself. We conclude that accurate best obtained by matching |Zin| (subject speaker) windowing.