Paper reading and analysis Heart Rate Monitoring During Physical Exercise From Photoplethysmography Using Neural Network

main content:

1. Explore the problem of other noise effects after ACC removes motion artifacts;

2. Observe the correlation between ACC and heart rate changes and model;

3. Perform regression analysis on heart rate changes and remove abnormal results;

Algorithm framework:

1. Calculate $Sacc$：
$$S_{acc,i}=\sum _{j=1}^n(X_{acc,j}^2+Y_{acc,j}^2+Z_{acc,j}^2),$$

$$S_{acc}=S_{acc,i}-S_{acc,i-1}$$

2, calculation $b_i$:$1\times 6$向量，$[S_{acc,i-6},\dots ,S_{acc,i}]$；

3. NN predicts: $\hat{\mathbf{r}}$；

4. Estimated heart rate difference: $\mathbf{r}$；

Linear regression of the results of 5, 3 and 4;

6. Post-processing steps:

Filtering step:
$$h{r}^{\prime }={\left(I+\lambda D^{T}D\right)}^{-1}hr$$
if filtered out, fit:
$$f\left(x\right)=a{\left(x-x_1\right)}^3+b{\left(x-x_1\right)}^2+c{\left(x-x_1\right)}^1+d$$