This function implements an adaptive normalization method on a given respiratory signal.
adanorm(x, sRate)Numeric vector representing the input signal to be normalized.
Integer value representing the sampling rate of the signal (number of samples per second).
Numeric vector representing the adaptively normalized signal.
It is designed to preserve the parts of the signal where the amplitude of respiration is small, typically when the body maintains a sleeping posture for extended periods.
Adaptive normalization first segments the signal into 1 second window before computing \(A(k)\) and is based on the following equations:
Equation (1) - Mean absolute deviation: $$A(k) = \frac{1}{fs} \sum_{i=k \cdot fs}^{(k+1) \cdot fs - 1} \left| x(i) \right|$$
Equation (2) - Standard deviation: $$\sigma(k) = \sqrt{\frac{1}{fs-1} \sum_{i=k \cdot fs}^{(k+1) \cdot fs - 1} (x(i) - \bar{x}(k))^2}$$
Equation (3) - Adaptive normalization factor, initialized to 1. $$F_{\text{norm}}(k) = \min\left(0.95F_{\text{norm}}(k-1) + 0.05A(k), 0.95F_{\text{norm}}(k-1) + 0.05\sigma(k)\right)$$
Choi, S. H., Yoon, H., Kim, H. S., Kim, H. B., Kwon, H. B., Oh, S. M., Lee, Y. J., & Park, K. S. (2018). Real-time apnea-hypopnea event detection during sleep by convolutional neural networks. In Computers in Biology and Medicine (Vol. 100, pp. 123–131). Elsevier BV. https://doi.org/10.1016/j.compbiomed.2018.06.028