Design and Implementation of a Multifunction Wearable Device to Monitor Sleep Physiological Signals

Figures

• 

  Figure 1. Diagram of our physiological sleep signal monitoring system, including the hardware of the portable biosignal acquisition unit. The device used a sound sensor, light sensor, triple-axis accelerometer, and temperature sensor, which were all integrated in a Circuit Playground Express (CPE) board to detect motion, ambient light, noise, and temperature. The CPE board has a diameter of 50 mm and a thickness of 1.6 mm. The collected data were stored in the 2-MB flash storage of the CPE, and the device can store up to 12 months of data from all the sensors required for data analysis. A photoplethysmographic (PPG) optical sensor (green light, 570 nm) was also included for heartbeat detection. (A) The system as worn on a user’s wrist with PPG sensor worn on the finger. (B) Close up view of the CPE board and the various sensors.

• 

  Figure 2. The components and an assembled schematic diagram of our wearable device. (A) A case with wrist straps was designed to enclose the CPE board so that the device could be worn like a watch for sleep analysis. The 3D-printed case had a diameter of 58 mm and a thickness of 65 mm. (B,C) Schematic of the wrist straps that were 92 mm long and 65 mm thick. (D) Schematic of the case enclosing the CPE board.

• 

  Figure 3. Sensor data collected from a subject wearing the developed device. The data from the light sensor are shown in yellow, the motion data are indicated in black, and the heart rate (HR) data from the PPG optical sensor are shown in red.

• 

  Figure 4. A schematic diagram for an experiment comparing our device to ActiWatch2. The subject wore both our device and ActiWatch2 on the right arm for the same time period every day for a week to collect data from both devices.

• 

  Figure 5. Comparison of the data collected using our device and the ActiWatch2. The data collected from ActiWatch2 are shown in (A), and the data collected from our device are shown in (B). The ambient light data (shown in yellow) collected from both devices are similar. Peaks occur at similar time points, although the slopes differ. This discrepancy could have been caused by the differences in the photoresistors used and in the locations of the light sensors on the devices. The motion sensor data (shown in black) are also similar between the two devices. Peaks occur at similar times, although the amplitudes differ. This discrepancy could have been caused by the difference in the driving and output voltages.

• 

  Figure 6. Correlation analyses of the data collected using our device and the ActiWatch2. The motion detection data are compared in (A). Large peaks generally occurred at the same times for both the ActiWatch2 and our device. The slight differences can be explained by the difference in the sampling frequencies of the two devices. Ambient light detection is compared in (B). The differences in amplitude may have been caused by differences in the photoresistors used in the two devices. The motion detection data from ActiWatch2 and our device were shown to be correlated, with a correlation coefficient (r) of 0.78, as shown in (C). The light detection data from ActiWatch2 and our device were shown to be highly correlated, with r = 0.89, as shown in (D).

• 

  Figure 7. Sleep analysis was performed using MATLAB software. An example analysis of 1 week of data is shown. The data from each 24-h period are displayed separately. The motion detection data are displayed in black, the light detection data are in yellow, and the heartbeat data are in red. The blue box indicates the sleep duration, as determined from the algorithm using both motion detection and ambient light sensing data.

• 

  Figure 8. Comparison of the sleep duration analysis results using our device and using the Food and Drug Administration (FDA)-approved ActiWatch2. (A) The results of sleep analysis performed with ActiWatch2, with the blue sections representing the sleep duration. (B) The results of the sleep analysis performed based on the data collected by our device. The sleep duration is indicated by the blue line. For the experiment, each 24-h interval started at noon and ended at noon on the next day. Seven intervals were continuously recorded. The motion detection data are displayed in black, and the light detection data are displayed in yellow. The durations of sleep time recorded by ActiWatch2 from Monday to Sunday were 7.5, 9, 8.6, 8.3, 6, 7.3, and 7 h, and the durations of sleep time recorded using the proposed device were 7.7, 9, 8.1, 8.3, 6.2, 7.5, and 7.6 h. A comparison of (A) and (B) shows that there is strong agreement between the sleep analysis results from the two devices, with the exception of the interval circled in red, for which there was a false positive. This misjudgment may have been caused by the user adjusting the watch due to unfamiliarity with wearing it for the first time.