Table 1.
Overview of studies on health information measurement using different low-cost sensors.
| Reference | Participant | Sensors | Measurement | Evaluation |
|---|---|---|---|---|
| [19] | A male subject | Accelerometer, cardiac activity electrodes, and inductive plethysmographic sensor mounted on the T-shirt | Respiration rate, heart rate, and movement of the body | Absolute mean percent differences are heart rates < 2% and respiration rates < 5%. |
| [20] | N = 14 | Temperature sensor embedded in KN95 mask | Respiration rate detection is based on the measurement of temperature variations through the vent holes of the mask and results can be applied to COVID-19 prevention | MAE = 0.449 BPM |
| [21] |
N = 10 (M = 7, F = 3) Age = 23.8 ± 0.84 years Height = 173.1 ± 6.9 cm Body mass = 66.2 ± 6.9 kg |
IMU and pressure sensor attached to insole and on legs | Motion gait recognition | Accuracy = 99.96% |
| [22] |
N = 16 (M = 10, F = 6) Age = 20–54 years |
Temperature sensor, pulse oximetry sensor, accelerometer, and GSR sensor attached to the upper arm | Perspiration measurement, activity recognition, skin temperature, blood oxygen saturation, heart rate | Accuracy = 87.5% (activity recognition) |
| [23] | N = 16 | PPG and GSR sensors attached to fingertip and wrist | Stress index | Accuracy = 85.3% |
| [24] |
N = 10 (M = 7, F = 3) Age = 26 ± 3 years Height = 165 ± 8 cm Body mass = 60 ± 10 kg |
Accelerometer and angular velocity sensors attached to low back and leg | Postural detection | Statistical results show IMU sensors are suitable for detection and evaluation of anticipatory postural adjustments |
| [25] |
N = 12 (M = 7, F = 5) Age = 24.91 ± 2.74 years Height = 166.91 ± 6.76 cm BMI = 61.41 ± 8.69 |
Textile capacitive proximity sensor placed under the feet | Gait measurement | Error rate of stride <1% Correlation coefficient between the reference sensor and the textile sensor is 0.865 |
| [26] | N = 5 | ECG sensor, pulse oximeter, temperature sensor attached to fingertip and body | Heart rate, respiratory rate, blood oxygen saturation, and body temperature | Accuracy = 99.26% |
| [27] |
N = 25 (M = 10, F = 15) Averaged age = 56.25 years |
IMU sensors attached to wrist or elbow | Shoulder joint mobility | Correlation coefficients between IMU and the traditional method are 0.997, 0.978, 0.897, and 0.984 for flexion, abduction, external rotation, and internal rotation, respectively |
| [28] |
N = 12 Age = 18–57 years |
Six 3-axis accelerometers and 12 gyroscopes placed on neck, left wrist, right wrist, waist, left leg, and right leg | Posture recognition (sitting, standing, walking, and lying) | Accuracy = 99.72% |
| [29] |
N = 100 (M = 50, F = 50) Age = 33 ± 6 years (M) Age = 32 ± 9 years (F) BMI = 23.1 ± 2.7 (M) BMI = 20.5 ± 2.1 (F) |
10 plantar pressure sensors underneath the insole | Foot pressure values during walking and standing | Statistical results show no significant difference between men and women in centre of pressure, while women exhibit higher peak pressure on the hallux, toes, forefoot, and medial aspect of the foot |
| [30] | N = 5 | Wearable inertial sensors attached to upper arm and forearm | Therapeutic movement measurement guided by therapists aiming to recover after the motion impairment | Specificity = 100% Sensitivity = 100% |
| [31] |
N = 3 (M = 3, F = 0) Age = 24.7 ± 2.4 years Height = 174.3 ±4.2 cm Body mass = 65.3 ± 7.0 kg |
A triaxis accelerometer and three single-axis gyro sensors attached to left/right thigh and left/right shank | Angular velocity and acceleration | Hip joint angle (flexion–extension) RMSE = 8.72, ADE = 6.57, CC = 0.88, PVU = 20.05% Hip joint angle (abduction–adduction) RMSE = 4.96, ADE = 3.30, CC = 0.72, PVU = 39.29% Knee joint angle (flexion–extension) RMSE = 6.79, ADE = 4.65, CC = 0.92, PVU = 14.60% |
| [32] |
N = 116 Age = 69 ± 18 years BMI = 27 ± 6 |
A piezoelectric sensor under the mattress | Respiratory rate, heart rate, and motion level | Specificity = 93%. Sensitivity = 85% |
| [33] |
N = 30 (M = 21, F = 10, where 11 paraplegic and 19 tetraplegic subjects) Age = 46.43 ± 16.91 years |
3D accelerometer and 3D gyroscope attached on each wrist and the right wheel of the wheelchair | Acceleration and peak velocity | Accuracy = 90% |
| [34] | PD patients N = 48 (M = 25, F = 23) Age = 70.61 ± 9.51 years Healthy people N = 40 (M = 22, F = 18); Age = 69.36 ± 7.42 years |
Accelerometer attached to left/right knee | Gait characteristics of PD patients and healthy people | Specificity = 90.91%. Sensitivity = 92.86% Accuracy = 88.46% |
| [35] |
N = 28 (M = 22, F = 6) Age = 41 ± 12 years Height = 70 ± 4 inch Body mass = 175 ± 43 lb |
Eight force sensors placed on wheelchair cushions | Peak pressure index, weight shift frequency, pressure relief frequency, in-seat activity frequency | Statistical results show force sensors can effectively monitor wheelchair users’ movements |
| [36] |
N = 27 (M = 12, F = 15) Age = 50.24 ± 12.99 years (M) Age = 40.93 ± 10.27 years (F) Height = 174.88 ± 10.25 cm (M) Height = 160.53 ± 4.31 cm (F) Body mass = 79.03 ± 11.99 kg (M) Body mass = 58.23 ± 7.83 kg (F) |
IMU sensor clipped to the back of marathon runners’ shorts | Step frequency, change in forward velocity, vertical oscillation, side-to-side movement of the pelvis, side-to-side drop of the pelvis, ground contact time | Statistical analysis shows IMU-based biomechanical indices can be used to detect fatigue in marathon runners |
| [37] | N = 7 | Eight hetero-core fibre optic pressure sensors placed on bed cushion | Respiration rate | Sensitivity = 0.05–0.2 dB |
| [38] | N = 7 | Pulse oximeter and heart rate sensor, thermometer, and ECG sensor | Oxygen saturation (SpO2), heart rate, body temperature, ECG | Accuracy = 1.02% (blood oxygen saturation detection) Accuracy = 0.51% (body temperature measurement) |
| [39] | N = 8 | IMU attached to the arm | Measure shoulder and elbow joint angles to continuously monitor human movement | Average correlation coefficient is >0.95 between the inertial tracker and the optical reference system. RMSE < 8 º (averaged value of eight subjects for all tasks) Peak-to-peak error < 12 º |
| [40] |
N = 72 (M = 39, F = 33) BMI = 28.74 ± 4.99 (HFR) BMI = 28.7 ± 4.81 (LFR) Age = 71.87 ± 6.45 years (HFR) Age = 63.47 ± 8.74 years (LFR) |
Four inertial sensors attached above and below each knee | Completion times for each test subactivity, joint range of motion, and flexion/extension velocities and accelerations | Accuracy = 90% Sensitivity = 94% Specificity = 59% |
| [41] |
N = 5 (M = 1, F = 4) Age = 24.8 ± 3.5 years |
Conductive textile sensors For lateral bending measurement, sensors are placed on both sides under the angle of the mandible and in correspondence with the trapezius scapula insertion; For axial rotation motion, sensors are placed on both sides on the anterior part of angle of the mandible and in correspondence with the trapezius muscle; For flexion–extension movement, one sensor is placed between the hyoid bone and the sternum (extension), the other between C2 and C7 vertebrae (flexion) |
Measure the angle (in degrees) of lateral bending, rotation, and flexion–extension of cervical spine movement | RMSE values of lateral bending, axial rotation, and flexion/extension of neck were 6.04 ± 0.67, 10.16 ± 2.11, and 12.31 ± 3.22, respectively |
| [42] |
N = 13 (M = 13, F = 0) Age = 26.1 ± 2.9 years Height = 178.7 ± 5.5 cm Body mass = 78.4 ± 5.9 kg |
Two identical, custom-built, six-degrees-of-freedom IMUs (accelerometer and gyroscope) attached to the right thing and shank via a knee sleeve | Knee joint forces | Accuracy Vertical force: RMSE = 19.1% ± 4.0%, anterior–posterior: RMSE = 21.8% ± 2.6%, medial–lateral: RMSE = 38.0% ± 6.1% |
| [43] | Young subjects: N = 21 Aged = 28.3 ± 6.8 years Body mass = 67.2 ± 9.6 kg Height = 1.70 ± 0.04 m Fallers: N = 16 Aged = 67.2 ± 6.7 years Body mass = 64.3 ± 12.0 kg Height = 1.58 ± 0.07 m |
Four load cells fixed to the chair | Force between sitting and standing swap | Error < 10% |
| [44] | N = 6 (five aged 22 to 23 and one aged 60) | Two dual-axis accelerometers orthogonally mounted on the waist | Daily activity detection | Accuracy = 90.8% (12 tasks) Accuracy = 94.1% (postural recognition) Accuracy = 83.3% (walking recognition) Accuracy = 95.6% (falling detection) |
| [45] |
N = 2 (M = 1, F = 1) Female: 1.58 m in height, 53 kg in body mass, 25 years; Male: 1.77 m in height, 75 kg in body mass, 24 years |
Three wireless transceiver modules fixed to the arm/leg with an elastic band | Arm and leg movements | Matching rate using two features: 70% and 80% for females and males, respectively. Matching rate using five features: 90% and 100% for females and males, respectively. |
| [46] |
N = 8 Age = 20–35 years |
48 fibre-optic pressure sensors placed below the mattress | Breathing rate, torso movement, sleep monitoring | Sensitivity = 71% Specificity = 87% |
| [47] |
N = 9 (M = 3, F = 6) Age = 23.3 ± 2.5 years, Body mass = 55.4 ± 8.5 kg, Height = 1.60 ± 0.08 m |
One triaxial accelerometer and three uniaxial gyroscopes were secured onto the back of the subjects | Angular measurements during trunk movement; trunk postural change | Correlation coefficients between the Vicon video capture system and sensors: >0.994 for dynamic tilting measurements and >0.776 for trunk postural measurements |
| [48] | N = 8 (four healthy subjects and four stroke survivors) | Triaxial accelerometers and triaxial gyroscopes worn on waist | Arm movement | Healthy people: Accuracy = 86% (accelerometer) and 72% (gyroscope) Stroke patients: Accuracy = 67% (accelerometer) and 60% (gyroscope) |
| [49] |
N = 8 (M = 2, F = 6) Age = 30 ± 5 years Body mass = 70 ± 15 kg |
IMU sensors placed on a glove worn by the driver | Stress indicators: emergency braking and rapid turning | Accuracy = 94.78% |
| [50] |
N = 17 (M = 8, F = 9) Age = 21.9 ± 3.7 years |
IMU sensors attached the right leg to Velcro strap | Knee flexion/extension angles | RMSE = 5.0º ± 1.0º MAE = 3.9º ± 0.8º |
| [51] |
N = 70 Age = 18–86 years |
IMU, temperature, pressure and GSR sensor attached to thigh with Velcro strap | Acceleration, angular velocity, skin temperature, muscle pressure, and sweat rate | The concordance correlation coefficient is 0.96 in comparison with the video motion analysis system. The highest estimation error for stride length was 4.81 cm (3.3%), and the mean error (N = 10) was 2.48 cm (1.7%). For gait speed, the estimation error < 3.8% (5.10 cm/s) and the mean error was 2.1%. |
| [52] |
N = 8 (M = 5, F = 3) Age = 21–24 years |
Piezoelectric sensor fixed to arms by bandage | Hand and wrist movements | Accuracy = 96.1% (LDA) Accuracy = 94.8% (ANN) |
| [53] |
N = 10 Age = 21–36 years Height = 1.48–1.89 m Body mass = 46.7–91.0 kg |
Five FSR sensors attach to the foot surface | Ground reaction forces | The correlations between FSR-based system and the gold standard force plate are 0.74–0.84. For hopping, the maximum GRF difference between FSR-based system and the gold standard force plate ranged from −6% to +14%. |
| [54] |
N = 5 (M = 5, F = 0) Age = 31 ± 5 years Height = 170 ± 4.6 cm Body mass = 71.2 ± 4.2 kg |
7 × 5 conductive textile sensors attached to leg | Knee angle during flexion–extension movements | MAE = 17.54 º RMSE = 18.82 º |
| [55] |
N = 7 (M = 4, F = 3) Age = 21–60 years |
64 × 128 pressure-sensitive e-textile sensors placed on bed | Respiration rate, leg movement | Precision = 70.3% Recall = 71.1% |
| [56] |
N = 8 Age = 25 ± 3 years Body mass = 61 ± 19 kg |
Triaxial accelerometer worn on the body | Falling detection | Sensitivity = 100% Specificity= 100% |
| [57] |
N = 8 (M = 4, F = 4) Age = 23.6 ± 1.3 years Height = 1.69 ± 0.08 m Body mass = 56.2 ± 10.3 kg |
Temperature sensor arrays fixed on the seat | Body–seat interface temperature measurement | Temperature field at the contact surface was not uniformly distributed. Heating rates = 1.7 ± 0.4 °C/min (fabric cover + foam) Heating rates = 1.6 ± 0.2 °C/min (wood) Heating rates = 1.7 ± 0.2 °C/min (leatherette cover + foam) |
| [58] |
N = 78 (M = 39, F = 39) Participants equally divided into three groups (n = 26) Group 1 Age = 21.9 ± 1.8 years BMI = 21.6 ± 2.8 kg/m2 Group 2 Age = 22.5 ± 2.4 years BMI = 22.2 ± 3.8 kg/m2 Group 3 Age = 22.2 ± 3.8 years BMI = 21.7 ± 2.1 kg/m2 |
Four digital humidity and temperature sensors placed under the ischial tuberosities and thighs bilaterally | Skin temperature and relative humidity | Temperature difference after two hours: 3.9 ± 1.4 °C (ischial tuberosities) and 5.6 ± 1.3 °C (thighs), 2.8 ± 1.7 °C (ischial tuberosities) and 4.7 ± 1.4 °C (thighs), 3.9 ± 1.3 °C (ischial tuberosities) and 6.3 ± 1.1 °C (thighs) for air-filled rubber, foam–fluid hybrid and medium density foam, respectively. No significant difference in relative humidity between different cushions |
| [59] |
N = 5 (F = 3, M = 2) Age = 33 ± 8 years Height = 180 ± 10 cm Body mass = 70 ± 21 kg |
Flexible screen-printed piezoresistive sensors | Four sitting posture recognition | Accuracy = 80%. |
| [60] |
N = 12 (M = 7, F = 5) Age = 22–36 years BMI = 16–34 kg/m2 |
FSR sensors (seven on seat pan and 5 on backrest) | Five sitting posture recognition | Accuracy = 96.85% |
| [61] |
N = 41 (M = 25, F = 16) Age = 24–64 years Height = 160–200 cm Body mass = 53–126 kg |
FSR sensors (10 on seat pan 4 on backrest, 2 on armrest) | Seven sitting posture recognition | Accuracy = 98% |
| [62] |
N = 9 Age = 59.7 ± 24.2 years Height = 1.76 ± 0.10 m Body mass = 38.78 ± 4.94 kg |
Customised piezoresistive sensors (eight sensors on seat pan and eight on backrest) | 12 sitting posture recognition | Repeatability and replicability of the system are evaluated. The total cost of the system is <150 USD in comparison to commercial products with a price of ~7000 USD. |
| [63] | N = 25 (M = 15, F = 10) | Customised fibre-based yarn coated with piezoelectric polymer placed on seat | Seven sitting posture recognition | Accuracy = 85.9% |
| [64] | N = 9 (M = 6, F = 3) | Customised textile pressure sensors placed on seat | 16 sitting posture recognition | Accuracy = 82% |
| [65] |
N = 36 (M = 21, F = 15) Age = 26.7 ± 2.0 years (M) Age = 25.0 ± 2.3 years (F) Height = 175.9 ± 6.4 cm (M) Height = 162.8 ± 4.6 cm (F) Body mass = 77.1 ± 15.0 kg (M) Body mass = 51.4 ± 4.3 kg (F) |
Six FSR sensors embedded in the seat cushion and six IRD sensors placed in the seatback | 11 sitting posture classification | Accuracy = 92% |
| [66] |
N = 8 (M = 8, F = 0) Age = 24–40 years |
Two IMU sensors placed on the lower and upper arms (near the wrist and elbow joints), respectively |
Movement of upper limbs | Angle error < 3º Position error < 9 mm |
| [67] |
N = 10 Age = 19–28 years Height = 155–187 cm Body mass = 46–70 kg |
Three RF sensors placed on the back of the subjects (thoracic, thoracolumbar, and lumbar regions) at the distance of 10 cm each | Sitting posture recognition | Accuracy = 98.83% |
| [68] |
N = 19 (M = 14, F = 5) Age = 22–58 years |
Two FSR sheets placed on seat pan (9 × 9) and backrest (10 × 9) | 15 sitting postures | Accuracy = 88.52% |