Linear and Non-Linear Heart Rate Variability Indexes from Heart-Induced Mechanical Signals Recorded with a Skin-Interfaced IMU

Figures

• 

  Figure 1. Schematic representation of the positioning of the sensors on the subject’s body and raw waveforms obtained from the ECG sensor sampled at 250 Hz (red waveform on the left) and the IMU sensor for both the accelerometer sensor and the gyroscope sampled at 120 Hz. For accelerometers and gyroscopes, all three axes are shown (x in blue, y in orange, and z in yellow). Regarding the IMU sensor, the x-axis is oriented from foot to head, the y-axis laterally from left to right, and the z-axis is oriented from back to chest.

• 

  Figure 2. 5 s of collected data from a volunteer in a lying position. First subplot: ECG waveform with detected R peaks. The R-R peaks distance allows calculating IBIRR,i. Second subplot: raw accelerations along the x-, y-, and z-axes. Third subplot: the obtained SCG signal with detected AO peaks. The AO-AO peaks distance allows calculating IBIAOAO,i. Fourth subplot: raw angular velocities along the x-, y-, and z-axes. Fifth subplot: the obtained GCG signal with detected AO peaks. The AO-AO peaks distance allow calculating IBIAOAO,i.

• 

  Figure 3. Example of 20 s of SCG and GCG signals (normalized between 0 and 1) obtained with algorithms 1 and 2 after the Hilbert filter application (blue lines) compared to raw SCG and GCG signals (orange lines) and reference ECG (black line, normalized between 0 and 1). In this specific case, Algorithm 2 selected the z-axis for SCG, while for GCG the same algorithm selected an axis different from y (default selection for Algorithm 1).

• 

  Figure 4. Bland-Altman analysis related to IBI comparison between IBIRR,i and (a) IBIAOAO,i estimated from SCG in seated position, (b) IBIAOAO,i estimated from GCG in seated position, (c) IBIAOAO,i estimated from SCG in lying position and (d) IBIAOAO,i estimated from GCG in lying position, for algorithm 1 (panels on the left) and algorithm 2 (panels on the right).

• 

  Figure 5. Percentage of selected axes used to retrieve the SCG signals and GCG signals from algorithm 2.

• 

  Figure 6. R values (between 0 and 1) obtained in the correlation analysis carried out between values retrieve from SCG and GCG against reference ECG values in both seated and lying position, for both the algorithms.

• 

  Figure 7. Percentage error between HRV derived from IBIAOAO estimated from SCG and GCG signals and from IBIRR from ECG signal. All the errors are reported in %.