BME 761

BME 761: Intelligent Biomedical Systems
Semester Project Proposal
Hand Gesture Recognition Using Wavelet Transforms
Done By Yousef Ahmad Abdallah
Uni ID 176806
Introduction
Hand Gesture is an Important tool for decoding sign language, which is
useful for communicating with the deaf and helps integrating them into
society. Another use for Hand Gesture Recognition is provide control
inputs to human Machine Interface for communication and control. Hand
Gesture can also be used to control prosthesis & control their force output
[1,2,3]. Throughout the years, many Input Modalities has been used for
hand gesture recognition, including Biopotentials [1,2,4], 2d Images [5],
Force Myography [6] & Doppler Shift [7].
Surface Electromyography is the recording of electrical signals generated

by muscles when they contract. It is a popular signal modality that allows
practitioners to measure many biomedical parameters including muscle
force, muscle degradation, limb angle and Muscle activation detection [8].
Since Hand Gestures are realised by different patterns of muscle
contraction, EMG is a highly valuable tool for inferring such muscle
contraction patterns. Several Works have implemented EMG as the signal
modality for their hand recognition systems [1,2,3]. Using Processed EMG
data can be useful for generating datasets that is specific for the patient
rather than relying on public sEMG Datasets that may not be suitable for
the patient due to differences between the patient and subjects involved
in the public dataset
Feature Extraction of sEMG Many feature extraction techniques have been

proposed Including Principle Component Analysis [1], Short time Fourier
Transform [9], Wavelet Transform [2] & Empirical Mode Decomposition
[10]. All these techniques are used for capturing frequency & temporal
based features which could aid in the classification process.
While there has been systems that relies on signal processing techniques
alone. These were proven to be unreliable and sensitive to individual
differences. Hand Gesture Recognition is a pattern recognition problem, it
is paramount to utilize pattern recognition techniques (Machine Learning
techniques) for such applications which include supervised classifiers like
KNN, Liner Regression Models & Artificial Neural Networks.
Literature Review
Wang et al. [1] has developed a Human Computer Interface to control a
robotic car using hand gestures. They utilized EMG Hardware to collect
raw EMG from 5 subjects for 30 minutes, preprocess it using hardware &
software filtration techniques, and apply it to a supervised machine
learning classification model. The goal of the system is to decode one of
the five gestures the system is trained to recognize. 1D Convolutional
neural networks, Autoencoder Nural Networks were used in compared
with other primitive machine learning techniques like linear discriminant
analysis and SVM. Accuracy was used as measure performance of the
classifiers and the best accuracy of 63% percent has been yielded by the
1D Convolutional Neural Network.
De Andrade et al. [2] has developed a hand recognition system based on

machine learning techniques with discrete wavelet transform used for
feature extraction. They used autoencoder neural network and kNN
technique and obtained an accuracy of 88%.
Manar et al [11] Utilized recurrent neural networks for hand gesture

recognition to recognize Arabic Sign Language (ArSL). Using a public
sEMG dataset their Model has achieved an accuracy of 95.1%
Proposal
For this Course Semester Project, A Hand Gesture Recognition System
based on surface Electromyography will be developed. Signal Processing
& Supervised Machine Learning Techniques will be implemented to
recognise the hand gestures via the EMG Channels.
The Targeted Hand Gestures to be Recognized are shown in figure 1. They

were recommended as these gestures are common in sign language and
can be used in combination with each other to form words.
The Gestures Names are as follows:
(a) Cylindrical: Gesture for holding cylindrical tools.
(b) Tip: Gesture for holding small objects.
(c) Hook: Gesture for holding a handle.
(d) Palmar: Gesture for grasping with palm facing the object.
(e) Spherical: Gesture for holding spherical tools.
(f) Lateral: Gesture for holding thin, flat objects.

Figure 1
Classifier
sEMG Feature Training &
Evaluation &
Collection Extraction Classification
Validation
Figure 2
The System Architecture is shown in figure 2, and it will be described as

follows
1] Data Collection & Data Generation: sEMG Data will be collected from a
few individuals ranging from 4 to 10 individuals performing at least five
rounds of aforementioned six hand gestures, using sEMG Data Acquisition
Hardware as described in [8].
2] Feature Extraction & Selection: In this work, temporal & frequency

features will be extracted via discrete wavelet transform as they are less
iterative process and can be used for online decomposition of the sEMG.
The energy & Standard deviation of the signal components will be
calculate. For temporal Parameters, variance, skewness, zero crossings
kurtosis will be calculated & counted.
3] Training & Classification: Multiple classifiers will be used including

neural Networks, linear discriminant analysis , support vector machines &
K-nearest neighbour. 80% Precent of the sEMG Data will be used for
training and the rest will be used for testing. The models will be
implemented in MATLAB
4] Model evaluation & validation: Classification accuracy, sensitivity,

precession will be calculated and visualized using confusion matrices.
These results will be used to compare between the different classifiers and
show which one is best used with discrete wavelet transform for hand
gesture recognition.
References
[1] Wang, C., Wang, S., Zhao, X., Fang, J., & Zhang, M. (2021). EMG-Based
Gesture Recognition System and Its Hardware Implementation. IEEE
Xplore, 1, 420–428. https://doi.org/10.1109/icbase53849.2021.00084
[2] De Andrade, F. H. C., Pereira, F. G., Resende, C. Z., & Cavalieri, D. C.

(2019). Improving SEMG-Based hand gesture recognition using maximal
overlap discrete wavelet transform and an autoencoder neural network. In
IFMBE proceedings (pp. 271–279). https://doi.org/10.1007/978-981-13-
2517-5_42
[3] Lu, C., Qi, Q., Liu, Y., Li, D., Xian, W., Wang, Y., Chen, C., & Xu, X.
(2024). Exoskeleton Recognition of human movement intent based on
surface electromyographic signals: review. IEEE Access, 12, 53986–54004.
https://doi.org/10.1109/access.2024.3388044
[4] Phukan, N., Kakoty, N. M., Shivam, P., & Gan, J. Q. (2019). Finger
movements recognition using minimally redundant features of wavelet
denoised EMG. Health and Technology, 9(4), 579–593.
https://doi.org/10.1007/s12553-019-00338-z
[5] Pisharady, Pramod & Vadakkepat, Prahlad & Loh, A.P.. (2010). Graph
matching based hand posture recognition using neuro-biologically inspired
features. 11th International Conference on Control, Automation, Robotics
and Vision, ICARCV 2010. 1151-1156. 10.1109/ICARCV.2010.5707352.
[6] Al-Hammouri, S., Barioul, R., Lweesy, K., Ibbini, M., & Kanoun, O.
(2024). A wrapper framework for feature selection and ELM weights
optimization for FMG-based sign recognition. Computers in Biology and
Medicine, 179, 108817.
https://doi.org/10.1016/j.compbiomed.2024.108817
[7] Bai, Y., Lu, L., Cheng, J., Liu, J., Chen, Y., & Yu, J. (2020). Acoustic-based
sensing and applications: A survey. Computer Networks, 181, 107447.
https://doi.org/10.1016/j.comnet.2020.107447
[8] Fortune, B. C., Pretty, C. G., Chatfield, L. T., McKenzie, L. R., & Hayes,
M. P. (2019b). Low-cost active electromyography. HardwareX, 6, e00085.
https://doi.org/10.1016/j.ohx.2019.e00085
[9] Tsenov, G., Zeghbib, A., Palis, F., Shoylev, N., Mladenov, V.: Neural
networks for online classification of hand and finger gestures using
surface EMG signals. In: 8th Seminar on Neural Network Applications in
Electrical Engineering, 2006, IEEE, 2006, pp. 167–171 (2006)
[10] Sapsanis, C., Georgoulas, G., Tzes, A., Lymberopoulos, D.: Improving
EMG based classification of basic hand gestures using EMD. In: 2013 35th
Annual International Conference of the IEEE Engineering in Medicine and
Biology Society (EMBC), pp. 5754– 5757, Osaka (2013)
[11] Manar Maraqa, Raed Abu-Zaiter. (2008, Aug.). “Recognition of Arabic

Sign Language (ArSL) Using Recurrent Neural Networks,” IEEE First
International Conference on the Applications of Digital Information and
Web Technologies, (ICADIWT 2008), pp. 478-48. Available: doi:
10.1109/ICADIWT.2008.4664396

BME 761

Uploaded by

Copyright:

Available Formats

BME 761

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

BME 761

Uploaded by

Copyright:

Available Formats

BME 761: Intelligent Biomedical Systems

Semester Project Proposal

Hand Gesture Recognition Using Wavelet Transforms

Done By Yousef Ahmad Abdallah

Surface Electromyography is the recording of electrical signals generated

Feature Extraction of sEMG Many feature extraction techniques have been

De Andrade et al. [2] has developed a hand recognition system based on

Manar et al [11] Utilized recurrent neural networks for hand gesture

The Targeted Hand Gestures to be Recognized are shown in figure 1. They

The Gestures Names are as follows:

(a) Cylindrical: Gesture for holding cylindrical tools.

(b) Tip: Gesture for holding small objects.

(c) Hook: Gesture for holding a handle.

(e) Spherical: Gesture for holding spherical tools.

(f) Lateral: Gesture for holding thin, flat objects.

The System Architecture is shown in figure 2, and it will be described as

2] Feature Extraction & Selection: In this work, temporal & frequency

3] Training & Classification: Multiple classifiers will be used including

4] Model evaluation & validation: Classification accuracy, sensitivity,

[2] De Andrade, F. H. C., Pereira, F. G., Resende, C. Z., & Cavalieri, D. C.

[11] Manar Maraqa, Raed Abu-Zaiter. (2008, Aug.). “Recognition of Arabic

You might also like