The mHealth app was developed to measure and record rotational data in real time by using an Android smartphone’s motion sensors. Figures 1-7 show the mHealth app in use to gather the sensor data. The app was programmed for Android mobile platforms with software development kits greater than 21 using Android Studio [19]. This research used the app on the default settings of the Samsung Galaxy S8 [20] with Android 7.0 mobile operating system. The code is written in Java using the Android application programming interface [21]. This app uses the Android sensor framework to access sensor data as part of the hardware package that consists of 3 classes and 1 interface. The classes are SensorManager, Sensor, and SensorEvent. SensorManager accesses the device’s sensors, Sensor obtains the list of available sensors, and SensorEvent creates the sensor object that includes the raw sensor data. The interface, SensorEventListener, receives notifications when a sensor value or accuracy changes [22].

Sensors used in this app include the gyroscope, accelerometer, gravity, and magnetic sensors. The gravity and magnetic sensors are used to calculate the rotation matrix using the getRotationMatrix method, which belongs to the SensorManager class [23]. Raw sensor data visualizations are created using the Androidplot library [24]. The collected sensor data are stored in an SQLite database [25]. Two comma-separated value files are created from the saved participant information and sensor values. These files are saved on the device for further data analysis. Figure 1 shows the main fragment in the mHealth app. In the main fragment, the mHealth obtains the participant’s status such as any pain (back, leg, head, etc), any medication in the last 3 days, any problems walking, concussion experience, gender, race/ethnicity, height, and weight. After touching the “SUBMIT” button in the main fragment, the mHealth app proceeds to the fragment shown in Figure 2. After selecting the “START” button (Figure 2), mHealth changes the fragment to Stop Fragment, as shown in Figure 3. Then, the participant walks straight forward for 2-3 minutes and then returns to the starting location. After the participant returns to the starting location, the participant touches the “STOP” button, as shown in Figure 3. The walking data are recorded between the time of selecting the “START” button (Figure 2) and the time of selecting the “STOP” button (Figure 3). After touching the “STOP” button, an administrator opens the widget in the mHealth app (Figure 4) to save the data by swiping from left to right on the smartphone screen. The “Admin” button (Figure 4) is used to save the recorded walking data. When selecting the “Admin” option (Figure 4), the mHealth app moves to the Save Fragment. There is 1 checkbox (Figure 5) to confirm whether the walking data are valid or not. The checkbox is for indicating whether the test was invalidated by an interruption or an unexpected event during recording. By selecting the “SAVE” button (Figure 4), the data are saved to an SQLite database and a comma-separated value file on the local disk. The mHealth app has several functions to analyze the raw data in real time. By selecting functions such as “Metrics,” “Accelerometer,” and “Gyroscope” (Figure 4), the mHealth app displays the current status of the raw data, as shown in Figure 6 and Figure 7. Figure 6 shows the current accelerometer, gyroscope, and rotation matrix in real time. Figure 7 shows the graph of the sensor data.

The mHealth app was tested with 3 participants. Two participants were healthy (1 female and 1 male) and they did not have any problems walking; these 2 participants were in their twenties. The third participant with back pain was in his thirties and he had little pain in his back; however, he had no problem walking and we could not find any difference in his walking compared to the other participants. During the experiment, all the participants wore the smartphone in a pocket of a waistband at the center of their body, as shown in Figure 8. The smartphone was placed on the waistline with its top frontside facing the right side of the body. The participants were instructed to walk straight at their own pace in an indoor hallway of a building. They walked a distance of about 400 feet. They walked for about 2-3 minutes in a straight line and then returned to the starting location. The duration of the test time for each participant was 4-5 minutes. They were allowed only a single trial for the walking movement test.

The smartphone stored the motion sensor data in the real-time mHealth app on the phone. The app collected the sensor data in 20 milliseconds. The sensor data such as accelerometer (x, y, and z), gyroscope (x, y, and z), and rotation matrix (3×3) were stored in an SQLite database and a comma-separated value file. Data corresponding to the first 10 seconds and the last 10 seconds of each participant session were removed to account for the press time of the start and stop button activation within the mHealth app. Among the sensor data, the rotation matrix was used to determine the walking balance. Because the rotation matrices can represent the rotation of the origin frame into the reference frame, the rotation matrix is commonly used to measure posture balance [26-31]. The next section explains the analysis method using the rotation matrix.

Considering the differentiation in the balance control of the participants while walking, the rotation vector data are the most effective [32,33]. Therefore, the rotation vector was extrapolated from the rotation matrix data recorded with the smartphone. In the mHealth app, the xyz-axis rotation vectors of the middle of the waistline were identified as the center of gravity of each participant. The x-axis represents the body motion angle between the right and left side movements of a participant. The y-axis represents the body motion angle between the up and down movements of a participant. The z-axis represents the body motion angle between the forward and backward movements of a participant (Figure 8). The xyz-axis rotation vector was obtained from the rotation matrix. A rotation matrix is a matrix that is used to perform a rotation in Euclidean space. The Euler’s theorem on the axis of a three-dimensional rotation is formulated as follows:

In this study, we developed an mHealth app to measure the body movements of 3 participants wearing the smartphone on their waists while walking. The motion sensors of the smartphone were used to measure the walking movements and to analyze the rotation matrix data with the proposed method, which shows the variables of walking over time. The difference in each walking step of the 3 participants was compared using a formula (equation 4). Table 1 shows the statistical information of the walking movement data after applying the formula. The healthy male participant had the smallest standard deviation (Table 1), indicating that he had the most balanced walking movement. The healthy female participant showed symmetrical walking movement in the left-to-right (x-axis) and up-and-down (y-axis) movements in the variations of the x-y axis rotation vector (Figure 9), indicating that she had lesser bias in gait than the other two participants. The x-axis standard deviation in the healthy female participant was the highest because female hip is bigger than male hip [36]. The standard deviation of y-axis and z-axis of the participant with back pain was the highest (Table 1), indicating that he had a wider movement than the other two healthy participants. Thus, the participant with back pain had a lesser balanced walking movement while walking than the other participants. Our app showed that the participant with little back pain has a wide range of movement during walking. Specifically, this participant appeared to show more variations in the right-to-left and forward-to-backward movements.

This study has several limitations. The mHealth app was tested only in 3 participants. Nevertheless, the proposed mHealth app can effectively capture differences in the postural control during walking between healthy individuals and individuals with back pain. Further, our mHealth app is not available currently for observing the data in real time while walking. Therefore, we will implement a program that will make it possible to observe the data while walking.

In future research, we will conduct feasibility and efficacy testing with large pools of individuals with reduced physical mobility. For the tests, we are planning to collect walking data from high school students because we believe that this younger group is ideal for researching on walking data. This younger sample is lesser affected by aging/diseases related to walking than an older group. Further, high school students will be in school for 2 or 3 years before graduation, which is a long period for tracking their walking movements. After we collect enough walking movement data, we will compare our findings with the results from previous mHealth studies. In addition, any variable that may affect postural balance during walking such as body habitus (ie, slim vs obese) or level of daily physical activities will be considered in data collection and analysis with various sensors in smartphones. We will also consider evaluating the impact of the app at the health system level by using outcomes such as health care utilization and medication use. This study is beneficial since it provides a useful method for medical evaluation in rehabilitation and physical fitness and a means for participants to maintain a state of well-being. This research can be used to classify walking movements between people who have walking-related diseases and normal people. This classification can help in diagnosing their diseases. Thereafter, we will research the classification of walking movements based on diseases that affect walking.

Our study shows that mHealth and the walking rotation vector can be used to define the body walking movements. Currently, the most widely used assessments for measuring postural control are laboratory-based such as the previous walking movement research [12,13] and human balance research [2,3]. Along with requiring specialized equipment, such assessments typically do not provide real-time feedback. However, our proposed mHealth app and analysis methods support home-based measurements. The findings of this study support our mHealth app as a low-cost and easy-to-use alternative with minimal equipment required that provides sensitive walking balance assessment [6].