Hand Exoskeleton Development Based on Voice Recognition Using Embedded Machine Learning on Raspberry Pi

Abstract

The choice of using speech to control the exoskeleton is based on the number of exoskeletons that are controlled using the EMG signal, where the EMG signal itself has the weakness of the complexity of the signal which is influenced by the position of the electrodes, as well as muscle fatigue. The purpose of this research is to develop an exoskeleton device using voice control based on embedded machine learning on a Raspberry Pi minicomputer. In this study, two feature extraction types namely mel-frequency cepstral coefficient (MFCC) and zero-crossing (ZC), and two machine learning algorithms, namely K-nearest Neighbor (K-NN) and Decision Tree (DT) were evaluated. The hand exoskeleton development consists of 3D hand design, microphone, Raspberry Pi 4B+, PCA9685 servo driver, and servo motor. Microphone was used to record voice commands given. After model evaluation, it was found that the MFCC extraction combined with the K-NN algorithm and the best accuracy (96±7.0%). In the implementation, we found that the accuracy is 79±14.46% and 90±14.14% for open and close commands.

The role of machine learning in robotics

“…A further technique being studied is voice actuation. As an example, an Indonesian group recently reported trials involving a voice-actuated robotic hand, based on ML embedded on a Raspberry Pi minicomputer (Triwiyanto et al , 2022). Two feature extraction techniques and two ML algorithms were evaluated, and trials showed that the mel-frequency cepstral coefficient combined with the k-nearest neighbour ML algorithm gave the most accurate results: 79 ± 14.46% and 90 ± 14.14% for open and close hand commands, respectively.…”

A Perspective on Prosthetic Hands Control: From the Brain to the Hand

“…Incremental learning techniques like online, transfer, and lifelong learning have been utilized to enhance the sEMG-based control system with new knowledge and skills, without the need to retrain the entire model from the beginning [139]. By utilizing techniques that adapt successful models from previous subjects, the training time and effort required to control an upper limb prosthesis can be reduced.…”

“…By utilizing techniques that adapt successful models from previous subjects, the training time and effort required to control an upper limb prosthesis can be reduced. Additionally, these techniques can allow the user to learn new gestures or movements with the prosthetic hand without interfering with pre-existing ones [138,139].…”

“…The human brain can use memory consolidation to store and retrieve new information by transferring it from shortterm to long-term memory [10]. This allows it to learn new gestures or motions without affecting existing ones by forming new neural associations and representations [137,139]. Incremental learning techniques can use online learning, transfer learning, and lifelong learning methods to update and expand learning models with new data or tasks [10].…”

“…The human brain has the ability to overcome limitations in capacity and lifespan by using strategies like chunking, abstraction, and analogy [173]. On the other hand, incremental learning techniques must utilize methods like curriculum learning, metalearning, and multitask learning to tackle challenges related to scalability and generalization [139].…”