The objective of this work is to set up a methodology that considers missing data from a connected heartbeat sensor in order to propose a good replacement methodology in the context of heart rate variability (HRV) computation. The framework is a research project, which aims to build a system that can measure stress and other factors influencing the onset and development of heart disease. The research encompasses studying existing methods, and improving them by use of experimental data from case study that describe the participant’s everyday life. We conduct a study to modelize stress from the HRV signal, which is extracted from a heart rate monitor belt connected to a smart watch. This paper describes data recording procedure and data imputation methodology. Missing data is a topic that has been discussed by several authors. The manuscript explains why we choose spline interpolation for data values imputation. We implement a random suppression data procedure and simulate removed data. After that, we implement several algorithms and choose the best one for our case study based on the mean square error.

In this paper the capacity of non-uniform sampling rate conversion techniques, involving different interpolation methods, aimed at wow defect reduction, is examined. Involved are: linear interpolation, four polynomial-based interpolation methods and the windowed sincbased method. The examined polynomial methods are: Lagrange interpolation, polynomial fitting with additional noise reduction, Hermitan and Spline. The performance of an artificially distorted audio signal, restored using non-uniform resampling, is evaluated on the basis of standard audio defect measurement criteria and compared for all of the aforementioned interpolation methods. The chosen defect descriptors are: total harmonic distortion, total harmonic distortion plus noise and signal to noise ratio.