Uneingeschränkter Zugang

Machine learning-enhanced gesture recognition through impedance signal analysis

Hoang Nhut Huynh
Hoang Nhut Huynh
, 
Quoc Tuan Nguyen Diep
Quoc Tuan Nguyen Diep
, 
Minh Quan Cao Dinh
Minh Quan Cao Dinh
, 
Anh Tu Tran
Anh Tu Tran
, 
Nguyen Chau Dang
Nguyen Chau Dang
, 
Thien Luan Phan
Thien Luan Phan
, 
Trung Nghia Tran
Trung Nghia Tran
 und 
Congo Tak Shing Ching
Congo Tak Shing Ching
   | 11. Juni 2024
Journal of Electrical Bioimpedance's Cover Image
Über diesen Artikel
Vorheriger Artikel
Nächster Artikel
Zitieren
Online veröffentlicht: 11. Juni 2024
Seitenbereich: 63 - 74
Eingereicht: 31. März 2024
DOI: https://doi.org/10.2478/joeb-2024-0007
Schlüsselwörter
© 2024 Hoang Nhut Huynh et al., published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.

Figure 1:

An overview of the gesture recognition system.
An overview of the gesture recognition system.

Figure 2:

Evaluation of five hand gestures designed for holding, grasping, and relaxing activities.
Evaluation of five hand gestures designed for holding, grasping, and relaxing activities.

Figure 3:

The precision impedance analyzer WK6632.
The precision impedance analyzer WK6632.

Figure 4:

The electrode for gesture recognition data acquisition.
The electrode for gesture recognition data acquisition.

Figure 5:

Electrode configuration for impedance data collection: using four electrodes (E1-E2, E1-E3, E1-E4, E2-E3, E2-E4, and E3-E4).
Electrode configuration for impedance data collection: using four electrodes (E1-E2, E1-E3, E1-E4, E2-E3, E2-E4, and E3-E4).

Figure 6:

Comparative SNR performance graph for five hand gestures (A, B, C, D, and E), with each gesture measured for 208 values (04 electrodes) and conducted with 200 repetitions.
Comparative SNR performance graph for five hand gestures (A, B, C, D, and E), with each gesture measured for 208 values (04 electrodes) and conducted with 200 repetitions.

Figure 7:

The measurements obtained from electrode pairs across various gestures for three volunteers with 5 gestures A, B, C, D, and E: (a) Person 1, (b) Person 2, (c) Person 3.
The measurements obtained from electrode pairs across various gestures for three volunteers with 5 gestures A, B, C, D, and E: (a) Person 1, (b) Person 2, (c) Person 3.

Figure 8:

Investigating PCA for gesture recognition with 2 dimensions.
Investigating PCA for gesture recognition with 2 dimensions.

Figure 9:

Investigating PCA for gesture recognition with three dimensions.
Investigating PCA for gesture recognition with three dimensions.

Figure 10:

Standard deviation values for sample data from 3 volunteers with labeled gestures A, B, C, D, and E for Person 1, Person 2, and Person 3.
Standard deviation values for sample data from 3 volunteers with labeled gestures A, B, C, D, and E for Person 1, Person 2, and Person 3.

Figure 11:

Confusion matrices demonstrating the classification accuracy of machine learning models for hand gesture recognition.
Confusion matrices demonstrating the classification accuracy of machine learning models for hand gesture recognition.

Performance evaluation of machine learning models.

Model Fold Accuracy Precision Recall F1-Score
KNN Fold-1 0.70 0.72 0.70 0.72
Fold-2 0.93 0.96 0.93 0.92
Fold-3 0.91 0.92 0.91 0.90
Fold-4 0.94 0.96 0.94 0.94
Fold-5 0.81 0.87 0.81 0.79
Mean 0.86 0.89 0.86 0.85
GBM Fold-1 0.94 0.96 0.94 0.94
Fold-2 0.93 0.94 0.93 0.92
Fold-3 0.76 0.63 0.76 0.68
Fold-4 0.95 0.95 0.95 0.95
Fold-5 0.70 0.79 0.70 0.69
Mean 0.86 0.86 0.86 0.84
Naive Bayes Fold-1 0.72 0.73 0.72 0.71
Fold-2 0.93 0.96 0.93 0.92
Fold-3 0.96 0.97 0.96 0.96
Fold-4 0.91 0.95 0.91 0.89
Fold-5 0.70 0.74 0.70 0.67
Mean 0.84 0.87 0.84 0.83
LR Fold-1 0.80 0.79 0.80 0.78
Fold-2 0.94 0.96 0.94 0.94
Fold-3 0.93 0.94 0.93 0.92
Fold-4 0.93 0.95 0.93 0.92
Fold-5 0.85 0.89 0.85 0.83
Mean 0.89 0.90 0.89 0.88
Random Forest Fold-1 0.74 0.74 0.74 0.73
Fold-2 0.93 0.96 0.93 0.92
Fold-3 0.96 0.97 0.96 0.96
Fold-4 0.93 0.96 0.93 0.92
Fold-5 0.81 0.89 0.81 0.79
Mean 0.87 0.90 0.87 0.86
SVM Fold-1 0.72 0.73 0.72 0.71
Fold-2 0.94 0.96 0.94 0.94
Fold-3 0.94 0.95 0.94 0.94
Fold-4 0.94 0.96 0.94 0.94
Fold-5 0.80 0.86 0.80 0.78
Mean 0.87 0.89 0.87 0.86
eISSN:
1891-5469
Sprache:
Englisch
Zeitrahmen der Veröffentlichung:
Volume Open
Fachgebiete der Zeitschrift:
Technik, Bioingenieurwesen, Biomedizinische Elektronik, Biologie, Biophysik, Medizin, Biomedizinische Technik, Physik, Spektroskopie und Metrologie
Zeitschrift RSS Feed