Table 6.
Classified event, parameters/features and data source.
| Classified Event | Sensor Data | Parameters/Features | Classification Approach | Accuracy |
|---|---|---|---|---|
| Fall Detection | Accelerometer, Barometer[34] Accelerometer, Barometer [46] Accelerometer [42] Microphone [75] Accelerometer (as a vibration or impact sensor) [27] | SMV, SMA, Tilt angle, Differential pressure Magnitude of a moving-window standard deviation, Standard deviation of the vector magnitude, ratio of the polar angle calculated in consecutive windows of 20 samples, difference in the values of the polar angle in consecutive windows SMA, SMV, Orientation, Mel Freq. Cepstral Coefficients (MFCC), Sound event length, Sound event energy, Steered response power (SRP) (A12) Vibration event length, Vibration event energy, Shock response spectrum, | Threshold-based algorithm Not stated Threshold-based algorithm kNN Not stated |
1. 96.9% 2. 94.12% 3. 95.6% 4. 95% 5. 100% Sensitivity, And 100% Specificity |
| Gait Assessment and Fall Risk Estimation | GaitShoe [76]: accelerometer, bend sensor, gyroscope, Force Sensitive Resistor (FSR), Polyvinylidine Flouride stripe, and electric field sensor. | Stride length, Stride velocity (integration of acceleration); orientation; force distribution under foot, heel-strike timing, and toe-off; heel-strike timing, toe-off timing; Plantar flexion/dorsi-flexion, Flexion at metatarsals; Height of foot above ground | Not stated | Not stated |
| ADL Food Preparation&Feeding | RFID [24] | Object touch | Not stated | 81.2% |
| ADL Selfcare | RFID [24] Accelerometers, RFID [77] | Object touch Acceleration, Object touch | Not stated Proprietary algorithm |
1. 81.2% 2. 92.95% |
| ADL House Keeping | RFID [24] Accelerometers, RFID [77] | Object touch Acceleration, Object touch | Not stated Proprietary algorithm |
1. 81.2% 2. 100% |
| ADL (Ambulation, Transfer, Posture) | Accelerometer [38] | Averaged variance over three axes, RMS of signal derivative, mean of signal derivative, average entropy over three axes, average cross correlation between each two axes, average range over three axes, average main frequency of the FFT over three axes, total signal energy averaged over three axes, energy of 0.2 window around the main frequency over total FFT energy (three axes average), Averaged skewness over three axes, Averaged Kurtosis over three axes, Averaged range of cross covariance between each two axes, Averaged mean of cross covariance between each two axes. | k-Nearest Neighbour (kNN, k = 1 − 5, 7); BN with Gaussian priors | Not stated |
| ADL Communication and Leisure | 1. EOG [26] 2. Accelerometers, RFID [77] | 1. Sacade (mean, variance, max amplitude, etc.), Saccade duration, fixation (mean, variance, amplitude, etc.), Fixation (time between each two saccades) duration, Average blink rate blink (mean, variance, max amplitude, etc.). 2. Acceleration, Object touch | 1. SVM 2. Proprietary algorithm |
1. precision of 76.1% and recall of 70.5% 2. 93.02% |
| Energy Expenditure | Accelerometer [78] | Coefficient of Variation (CV) for six 10s epochs within a 1min period, Vo2, average CV and the average counts per minute were calculated for minutes 4–9 of each activity | Two-regression model | 95% |
| Location Determination | Orientation, tilt angle, | |||
| Subject Active or Inactive) | 1. Accelerometer 2. Accelerometer [42] |
1. Mean, SMA, Variance, STD. 2. SMA |
1. Not stated 2. Threshold-based algorithm |
1. 100% 2. 100% |
| Movement or Activity Intensity | Accelerometer [42,65] | Sample differences, Integral of RMS, Mean of Minmax, SMV, Cross correlation | ||
| Checking and Comparing Signals | Accelerometer [65] | Correlation coefficients, Sample differences, Signal Correlation, Cross correlation, Dynamic time warping |