Simultaneous Prediction of Wrist/Hand Motion via Wearable Ultrasound Sensing

Abstract

The ability to predict wrist and hand motions simultaneously is essential for natural controls of hand protheses. In this paper, we propose a novel method that includes subclass discriminant analysis (SDA) and principal component analysis for the simultaneous prediction of wrist rotation (pronation/supination) and finger gestures using wearable ultrasound. We tested the method on eight finger gestures with concurrent wrist rotations. Results showed that SDA was able to achieve accurate classification of both finger gestures and wrist rotations under dynamic wrist rotations. When grouping the wrist rotations into three subclasses, about 99.2 ±1.2% of finger gestures and 92.8 ± 1.4% of wrist rotations can be accurately classified. Moreover, we found that the first principal component (PC1) of the selected ultrasound features was linear to the wrist rotation angle regardless of finger gestures. We further used PC1 in an online tracking task for continuous wrist control and demonstrated that a wrist tracking precision (R 2 ) of 0.954 ± 0.012 and a finger gesture classification accuracy of 96.5 ± 1.7% can be simultaneously achieved, with only two minutes of user training. Our proposed simultaneous wrist/hand control scheme is training-efficient and robust, paving the way for musculaturedriven artificial hand control and rehabilitation treatment.Index Terms-Subclass discriminant analysis, principal component analysis, simultaneous wrist/hand control, wearable ultrasound sensing.

Motion prediction using electromyography and sonomyography for an individual with transhumeral limb loss

“…This high spatial specificity means that muscular cross-talk does not contaminate the extracted control signals that can be used to drive movement of a prosthesis. Numerous studies have established SMG as a viable option for gesture recognition and prosthesis control [23–26]. In particular, our group has demonstrated the ability to classify five individual digit movements with 97% accuracy [27] and 15 complex grasps with 91% accuracy [28] in able-bodied individuals.…”

Classification of Individual Finger Flexions Using Ultrasound Radiofrequency Signals

“…The optimal number and placement of the WUSs should be investigated to obtain the best classification accuracy. In addition, future studies should perform the same hand gestures and motions made in previous ultrasound transducer experiments [13,14,53,52,51] and sEMG studies [7,8,69,34]. Optimum parameters of the pattern classification pipeline were obtained by conducting three simultaneous independent experiments changing only the final classification step to be one of LDA, MLP or KNN classifier.…”

“…The dimensionality of the data matrix ∈ × is reduced to ∈ × with as the number of features, as the number of classes, and as the reduced dimension of features (where − features are removed). LDA is a computationally efficient and robust classification method that has been commonly used involving sEMG and ultrasound data for human machine interface applications [13,14,52,53,62].…”

“…Finally, every 100 ms of the finger flexion recordings are classified using linear discriminant analysis (LDA). LDA was used as the machine learning classifier because itis fast to compute and has been used in previous sEMG and ultrasound related studies[13,14,52,53,62].…”

“…Their feature extraction method was adapted by the two-dimensional (2D) linear regression method used by Castellini et al for the ultrasound RF signals[41]. Design of a wearable multichannel UT prototype sparked further researchers to pursue lower resolution designs for the purpose of removing the bulky clinical B-mode ultrasonic imaging probe with a more costeffective solution[52,53].…”