Optimizing the performance of convolutional neural network for enhanced gesture recognition using sEMG.

Deep neural networks (DNNs) have demonstrated higher performance results when compared to traditional approaches for implementing robust myoelectric control (MEC) systems. However, the delay induced by optimising a MEC remains a concern for real-time applications. As a result, an optimised DNN architecture based on fine-tuned hyperparameters is required. This study investigates the optimal configuration of convolutional neural network (CNN)-based MEC by proposing an effective data segmentation technique and a generalised set of hyperparameters. Firstly, two segmentation strategies (disjoint and overlap) and various segment and overlap sizes were studied to optimise segmentation parameters. Secondly, to address the challenge of optimising the hyperparameters of a DNN-based MEC system, the problem has been abstracted as an optimisation problem, and Bayesian optimisation has been used to solve it. From 20 healthy people, ten surface electromyography (sEMG) grasping movements abstracted from daily life were chosen as the target gesture set. With an ideal segment size of 200 ms and an overlap size of 80%, the results show that the overlap segmentation technique outperforms the disjoint segmentation technique (p-value < 0.05). In comparison to manual (12.76 ± 4.66), grid (0.10 ± 0.03), and random (0.12 ± 0.05) search hyperparameters optimisation strategies, the proposed optimisation technique resulted in a mean classification error rate (CER) of 0.08 ± 0.03 across all subjects. In addition, a generalised CNN architecture with an optimal set of hyperparameters is proposed. When tested separately on all individuals, the single generalised CNN architecture produced an overall CER of 0.09 ± 0.03. This study's significance lies in its contribution to the field of EMG signal processing by demonstrating the superiority of the overlap segmentation technique, optimizing CNN hyperparameters through Bayesian optimization, and offering practical insights for improving prosthetic control and human-computer interfaces.

Smart ECG Biosensor Design with an Improved ANN Performance Based on the Taguchi Optimizer.

This paper aims to design a smart biosensor to predict electrocardiogram (ECG) signals in a specific auscultation site from other ECG signals measured from other measurement sites. The proposed design is based on a hybrid architecture using the Artificial Neural Networks (ANNs) model and Taguchi optimizer to avoid the ANN issues related to hyperparameters and to improve its accuracy. The proposed approach aims to optimize the number and type of inputs to be considered for the ANN model. Indeed, different combinations are considered in order to find the optimal input combination for the best prediction quality. By identifying the factors that influence a model's prediction and their degree of importance via the modified Taguchi optimizer, the developed biosensor improves the prediction accuracy of ECG signals collected from different auscultation sites compared to the ANN-based biosensor. Based on an actual database, the simulation results show that this improvement is significant; it can reach more than 94% accuracy.

The Effect of Signal Duration on the Classification of Heart Sounds: A Deep Learning Approach.

Deep learning techniques are the future trend for designing heart sound classification methods, making conventional heart sound segmentation dispensable. However, despite using fixed signal duration for training, no study has assessed its effect on the final performance in detail. Therefore, this study aims at analysing the duration effect on the commonly used deep learning methods to provide insight for future studies in data processing, classifier, and feature selection. The results of this study revealed that (1) very short heart sound signal duration (1 s) weakens the performance of Recurrent Neural Networks (RNNs), whereas no apparent decrease in the tested Convolutional Neural Network (CNN) model was found. (2) RNN outperformed CNN using Mel-frequency cepstrum coefficients (MFCCs) as features. There was no difference between RNN models (LSTM, BiLSTM, GRU, or BiGRU). (3) Adding dynamic information (∆ and ∆²MFCCs) of the heart sound as a feature did not improve the RNNs' performance, and the improvement on CNN was also minimal (≤2.5% in MAcc). The findings provided a theoretical basis for further heart sound classification using deep learning techniques when selecting the input length.

Hammerstein-Wiener Multimodel Approach for Fast and Efficient Muscle Force Estimation from EMG Signals.

This paper develops a novel approach to characterise muscle force from electromyography (EMG) signals, which are the electric activities generated by muscles. Based on the nonlinear Hammerstein-Wiener model, the first part of this study outlines the estimation of different sub-models to mimic diverse force profiles. The second part fixes the appropriate sub-models of a multimodel library and computes the contribution of sub-models to estimate the desired force. Based on a pre-existing dataset, the obtained results show the effectiveness of the proposed approach to estimate muscle force from EMG signals with reasonable accuracy. The coefficient of determination ranges from 0.6568 to 0.9754 using the proposed method compared with a range of 0.5060 to 0.9329 using an artificial neural network (ANN), generating significantly different accuracy (p < 0.03). Results imply that the use of multimodel approach can improve the accuracy in proportional control of prostheses.

Software Sensor to Enhance Online Parametric Identification for Nonlinear Closed-Loop Systems for Robotic Applications.

This paper proposes an online direct closed-loop identification method based on a new dynamic sliding mode technique for robotic applications. The estimated parameters are obtained by minimizing the prediction error with respect to the vector of unknown parameters. The estimation step requires knowledge of the actual input and output of the system, as well as the successive estimate of the output derivatives. Therefore, a special robust differentiator based on higher-order sliding modes with a dynamic gain is defined. A proof of convergence is given for the robust differentiator. The dynamic parameters are estimated using the recursive least squares algorithm by the solution of a system model that is obtained from sampled positions along the closed-loop trajectory. An experimental validation is given for a 2 Degrees Of Freedom (2-DOF) robot manipulator, where direct and cross-validations are carried out. A comparative analysis is detailed to evaluate the algorithm's effectiveness and reliability. Its performance is demonstrated by a better-quality torque prediction compared to other differentiators recently proposed in the literature. The experimental results highlight that the differentiator design strongly influences the online parametric identification and, thus, the prediction of system input variables.

Altered evoked low-frequency connectivity from SI to ACC following nerve injury in rats.

Objective. Despite decades of research on central processing of pain, there are still several unanswered questions, in particular regarding the brain regions that may contribute to this alerting sensation. Since it is generally accepted that more than one cortical area is responsible for pain processing, there is an increasing focus on the interaction between areas known to be involved.Approach. In this study, we aimed to investigate the bidirectional information flow from the primary somatosensory cortex (SI) to the anterior cingulate cortex (ACC) in an animal model of neuropathic pain.19 rats (nine controls and ten intervention) had an intracortical electrode implanted with six pins in SI and six pins in ACC, and a cuff stimulation electrode around the sciatic nerve. The intervention rats were subjected to the spared nerve injury (SNI) after baseline recordings. Electrical stimulation at three intensities of both noxious and non-noxious stimulation was used to record electrically evoked cortical potentials. To investigate information flow, two connectivity measures were used: phase lag index (PLI) and granger prediction (GP). The rats were anesthetized during the entire study.Main results. Immediately after the intervention (<5 min after intervention), the high frequency (γandγ+) PLI was significantly decreased compared to controls. In the last recording cycle (3-4 h after intervention), the GP increased consistently in the intervention group. Peripheral nerve injury, as a model of neuropathic pain, resulted in an immediate decrease in information flow between SI and ACC, possibly due to decreased sensory input from the injured nerve. Hours after injury, the connectivity between SI and ACC increased, likely indicating hypersensitivity of this pathway.Significance. We have shown that both a directed and non-directed connectivity between SI and ACC approach can be used to show the acute changes resulting from the SNI model.

Online Closed-Loop Control Using Tactile Feedback Delivered Through Surface and Subdermal Electrotactile Stimulation.

AIM: Limb loss is a dramatic event with a devastating impact on a person's quality of life. Prostheses have been used to restore lost motor abilities and cosmetic appearance. Closing the loop between the prosthesis and the amputee by providing somatosensory feedback to the user might improve the performance, confidence of the amputee, and embodiment of the prosthesis. Recently, a minimally invasive method, in which the electrodes are placed subdermally, was presented and psychometrically evaluated. The present study aimed to assess the quality of online control with subdermal stimulation and compare it to that achieved using surface stimulation (common benchmark) as well as to investigate the impact of training on the two modalities. METHODS: Ten able-bodied subjects performed a PC-based compensatory tracking task. The subjects employed a joystick to track a predefined pseudorandom trajectory using feedback on the momentary tracking error, which was conveyed via surface and subdermal electrotactile stimulation. The tracking performance was evaluated using the correlation coefficient (CORR), root mean square error (RMSE), and time delay between reference and generated trajectories. RESULTS: Both stimulation modalities resulted in good closed-loop control, and surface stimulation outperformed the subdermal approach. There was significant difference in CORR (86 vs 77%) and RMSE (0.23 vs 0.31) between surface and subdermal stimulation (all p < 0.05). The RMSE of the subdermal stimulation decreased significantly in the first few trials. CONCLUSION: Subdermal stimulation is a viable method to provide tactile feedback. The quality of online control is, however, somewhat worse compared to that achieved using surface stimulation. Nevertheless, due to minimal invasiveness, compactness, and power efficiency, the subdermal interface could be an attractive solution for the functional application in sensate prostheses.

Estimation of the Respiratory Rate from Localised ECG at Different Auscultation Sites.

The respiratory rate (RR) is a vital physiological parameter in prediagnosis and daily monitoring. It can be obtained indirectly from Electrocardiogram (ECG) signals using ECG-derived respiration (EDR) techniques. As part of the study in designing an early cardiac screening system, this work aimed to study whether the accuracy of ECG derived RR depends on the auscultation sites. Experiments were conducted on 12 healthy subjects to obtain simultaneous ECG (at auscultation sites and Lead I as reference) and respiration signals from a microphone close to the nostril. Four EDR algorithms were tested on the data to estimate RR in both the time and frequency domain. Results reveal that: (1) The location of the ECG electrodes between auscultation sites does not impact the estimation of RR, (2) baseline wander and amplitude modulation algorithms outperformed the frequency modulation and band-pass filter algorithms, (3) using frequency domain features to estimate RR can provide more accurate RR except when using the band-pass filter algorithm. These results pave the way for ECG-based RR estimation in miniaturised integrated cardiac screening device.

The Effect of Spinal Manipulation on the Electrophysiological and Metabolic Properties of the Tibialis Anterior Muscle.

There is growing evidence showing that spinal manipulation increases muscle strength in healthy individuals as well as in people with some musculoskeletal and neurological disorders. However, the underlying mechanism by which spinal manipulation changes muscle strength is less clear. This study aimed to assess the effects of a single spinal manipulation session on the electrophysiological and metabolic properties of the tibialis anterior (TA) muscle. Maximum voluntary contractions (MVC) of the ankle dorsiflexors, high-density electromyography (HDsEMG), intramuscular EMG, and near-infrared spectroscopy (NIRS) were recorded from the TA muscle in 25 participants with low level recurring spinal dysfunction using a randomized controlled crossover design. The following outcomes: motor unit discharge rate (MUDR), strength (force at MVC), muscle conduction velocity (CV), relative changes in oxy- and deoxyhemoglobin were assessed pre and post a spinal manipulation intervention and passive movement control. Repeated measures ANOVA was used to assess within and between-group differences. Following the spinal manipulation intervention, there was a significant increase in MVC (p = 0.02; avg 18.87 ± 28.35%) and a significant increase in CV in both the isometric steady-state (10% of MVC) contractions (p < 0.01; avg 22.11 ± 11.69%) and during the isometric ramp (10% of MVC) contractions (p < 0.01; avg 4.52 ± 4.58%) compared to the control intervention. There were no other significant findings. The observed TA strength and CV increase, without changes in MUDR, suggests that the strength changes observed following spinal manipulation are, in part, due to increased recruitment of larger, higher threshold motor units. Further research needs to investigate the longer term and potential functional effects of spinal manipulation in various patients who may benefit from improved muscle function and greater motor unit recruitment.

A comparative study of motion detection with FMG and sEMG methods for assistive applications.

INTRODUCTION: While surface-electromyography (sEMG) has been widely used in limb motion detection for the control of exoskeleton, there is an increasing interest to use forcemyography (FMG) method to detect motion. In this paper, we review the applications of two types of motion detection methods. Their performances were experimentally compared in day-to-day classification of forearm motions. The objective is to select a detection method suitable for motion assistance on a daily basis. METHODS: Comparisons of motion detection with FMG and sEMG were carried out considering classification accuracy (CA), repeatability and training scheme. For both methods, classification of motions was achieved through feed-forward neural network. Repeatability was evaluated on the basis of change in CA between days and also training schemes. RESULTS: The experiments shows that day-to-day CA with FMG can reach 84.9%, compared with a CA of 77.8% with sEMG, when the classifiers were trained only on the first day. Moreover, the CA with FMG can reach to 86.5%, comparable to CA of 84.1% with sEMG, if classifiers were trained daily. CONCLUSIONS: Results suggest that data recorded from FMG is more repeatable in day-to-day testing and therefore FMG-based methods can be more useful than sEMG-based methods for motion detection in applications where exoskeletons are used as needed on a daily basis.

A Multiday Evaluation of Real-Time Intramuscular EMG Usability with ANN.

Recent developments in implantable technology, such as high-density recordings, wireless transmission of signals to a prosthetic hand, may pave the way for intramuscular electromyography (iEMG)-based myoelectric control in the future. This study aimed to investigate the real-time control performance of iEMG over time. A novel protocol was developed to quantify the robustness of the real-time performance parameters. Intramuscular wires were used to record EMG signals, which were kept inside the muscles for five consecutive days. Tests were performed on multiple days using Fitts' law. Throughput, completion rate, path efficiency and overshoot were evaluated as performance metrics using three train/test strategies. Each train/test scheme was categorized on the basis of data quantity and the time difference between training and testing data. An artificial neural network (ANN) classifier was trained and tested on (i) data from the same day (WDT), (ii) data collected from the previous day and tested on present-day (BDT) and (iii) trained on all previous days including the present day and tested on present-day (CDT). It was found that the completion rate (91.6 ± 3.6%) of CDT was significantly better (p < 0.01) than BDT (74.02 ± 5.8%) and WDT (88.16 ± 3.6%). For BDT, on average, the first session of each day was significantly better (p < 0.01) than the second and third sessions for completion rate (77.9 ± 14.0%) and path efficiency (88.9 ± 16.9%). Subjects demonstrated the ability to achieve targets successfully with wire electrodes. Results also suggest that time variations in the iEMG signal can be catered by concatenating the data over several days. This scheme can be helpful in attaining stable and robust performance.

The Variability of Psychophysical Parameters Following Surface and Subdermal Stimulation: A Multiday Study in Amputees.

Electrotactile stimulation has been suggested as a modality for providing sensory feedback in upper limb prostheses. This study investigates the multiday variability of subdermal and surface stimulation. Electrical stimulation was delivered using either surface or fine wire electrodes placed right under the skin in eight amputees for seven consecutive days. The variability of psychophysical measurements, including detection threshold (DT), pain threshold (PT), dynamic range (DR), just noticeable difference (JND), Weber fraction (WF) and quality of evoked sensations, was evaluated using the coefficient of variation (CoV). In addition, the systematic change in the mean of the parameters across days was assessed in both stimulation modalities. In the case of DT, PT, DR, and perceived intensity at 100 Hz, the CoV of surface stimulation was significantly smaller than that of subdermal stimulation. Only PT showed a significant systematic change in the mean value across days for both modalities. The outcome of this study has implications for the choice of modality in delivering sensory feedback, though the significance of the quantified variability needs to be evaluated using usability tests with user feedback.

Classification of Movement Preparation Between Attended and Distracted Self-Paced Motor Tasks.

OBJECTIVE: Brain-computer interface (BCI) systems aim to control external devices by using brain signals. The performance of these systems is influenced by the user's mental state, such as attention. In this study, we classified two attention states to a target task (attended and distracted task level) while attention to the task is altered by one of three types of distractors. METHODS: A total of 27 participants were allocated into three experimental groups and exposed to one type of distractor. An attended condition that was the same across the three groups comprised only the main task execution (self-paced dorsiflexion) while the distracted condition was concurrent execution of the main task and an oddball task (dual-task condition). Electroencephalography signals were recorded from 28 electrodes to classify the two attention states of attended or distracted task conditions by extracting temporal and spectral features. RESULTS: The results showed that the ensemble classification accuracy using the combination of temporal and spectral features (spectro-temporal features, 82.3 ± 2.7%) was greater than using temporal (69 ± 2.2%) and spectral (80.3 ± 2.6%) features separately. The classification accuracy was computed using a combination of different channel locations, and it was demonstrated that a combination of parietal and centrally located channels was superior for classification of two attention states during movement preparation (parietal channels: 84.6 ± 1.3%, central and parietal channels: 87.2 ± 1.5%). CONCLUSION: It is possible to monitor the users' attention to the task for different types of distractors. SIGNIFICANCE: It has implications for online BCI systems where the requirement is for high accuracy of intention detection.

Multiday Evaluation of Techniques for EMG-Based Classification of Hand Motions.

Currently, most of the adopted myoelectric schemes for upper limb prostheses do not provide users with intuitive control. Higher accuracies have been reported using different classification algorithms but investigation on the reliability over time for these methods is very limited. In this study, we compared for the first time the longitudinal performance of selected state-of-the-art techniques for electromyography (EMG) based classification of hand motions. Experiments were conducted on ten able-bodied and six transradial amputees for seven continuous days. Linear discriminant analysis (LDA), artificial neural network (ANN), support vector machine (SVM), K-nearest neighbor (KNN), and decision trees (TREE) were compared. Comparative analysis showed that the ANN attained highest classification accuracy followed by LDA. Three-way repeated ANOVA test showed a significant difference (P < 0.001) between EMG types (surface, intramuscular, and combined), days (1-7), classifiers, and their interactions. Performance on the last day was significantly better (P < 0.05) than the first day for all classifiers and EMG types. Within-day, classification error (WCE) across all subject and days in ANN was: surface (9.12 ± 7.38%), intramuscular (11.86 ± 7.84%), and combined (6.11 ± 7.46%). The between-day analysis in a leave-one-day-out fashion showed that the ANN was the optimal classifier (surface (21.88 ± 4.14%), intramuscular (29.33 ± 2.58%), and combined (14.37 ± 3.10%). Results indicate that within day performances of classifiers may be similar but over time, it may lead to a substantially different outcome. Furthermore, training ANN on multiple days might allow capturing time-dependent variability in the EMG signals and thus minimizing the necessity for daily system recalibration.

On the robustness of real-time myoelectric control investigations: a multiday Fitts' law approach.

OBJECTIVE: Real-time myoelectric experimental protocol is considered as a means to quantify usability of myoelectric control schemes. While usability should be considered over time to assure clinical robustness, all real-time studies reported thus far are limited to a single session or day and thus the influence of time on real-time performance is still unexplored. In this study, the aim was to develop a novel experimental protocol to quantify the effect of time on real-time performance measures over multiple days using a Fitts' law approach. APPROACH: Four metrics: throughput, completion rate, path efficiency and overshoot, were assessed using three train-test strategies: (i) an artificial neural network (ANN) classifier was trained on data collected from the previous day and tested on present day (BDT) (ii) trained and tested on the same day (WDT) and (iii) trained on all previous days including present day and tested on present day (CDT) in a week-long experimental protocol. MAIN RESULTS: It was found that on average, the completion rate (98.37% ± 1.47%) of CDT was significantly better (P < 0.01) than that of BDT (86.25% ± 3.46%) and WDT (94.22% ± 2.74%). The throughput (0.40 ± 0.03 bits s-1) of CDT was significantly better (P = 0.001) than that of BDT (0.38 ± 0.03 bits s-1). Offline analysis showed a different trend due to the difference in the training strategies. SIGNIFICANCE: Results suggest that increasing the size of the training set over time can be beneficial to assure robust performance of the system over time.

Performance of Combined Surface and Intramuscular EMG for Classification of Hand Movements.

The surface EMG (sEMG) has been used as control source for upper limb prosthetics since decades. Previous studies suggested that intramuscular EMG showed promising results for upper limb prosthetics. This study investigates the strength of combined surface and intramuscular EMG (cEMG) for improved myoelectric control. Five able-bodied subjects and three transradial amputees were evaluated using offline classification error as performance metric. Six surface and intramuscular channels were recorded concurrently from each subject for seven consecutive days and Stacked sparse autoencoders (SSAE) and LDA classifiers were used for classification. As a control source, either sEMG channels were used or combined channels were used with reduced features using PCA. In the within session analysis, cEMG $( 2.21 \pm 1.19${%) outperformed the sEMG ($4.63 \pm 2.07${%) for both able-bodied and amputee subjects using SSAE. For between session analysis, cEMG outperformed the sEMG for both able-bodied and amputee subjects with percentage points difference of 7.93. These results imply cEMG can significantly improve the performance of pattern recognition based myoelectric control scheme for amputee subjects too and further improvement can be made by utilizing SSAE which show improved performance as compared to LDA.

Classification of Overt and Covert Speech for Near-Infrared Spectroscopy-Based Brain Computer Interface.

People suffering from neuromuscular disorders such as locked-in syndrome (LIS) are left in a paralyzed state with preserved awareness and cognition. In this study, it was hypothesized that changes in local hemodynamic activity, due to the activation of Broca's area during overt/covert speech, can be harnessed to create an intuitive Brain Computer Interface based on Near-Infrared Spectroscopy (NIRS). A 12-channel square template was used to cover inferior frontal gyrus and changes in hemoglobin concentration corresponding to six aloud (overtly) and six silently (covertly) spoken words were collected from eight healthy participants. An unsupervised feature extraction algorithm was implemented with an optimized support vector machine for classification. For all participants, when considering overt and covert classes regardless of words, classification accuracy of 92.88 ± 18.49% was achieved with oxy-hemoglobin (O2Hb) and 95.14 ± 5.39% with deoxy-hemoglobin (HHb) as a chromophore. For a six-active-class problem of overtly spoken words, 88.19 ± 7.12% accuracy was achieved for O2Hb and 78.82 ± 15.76% for HHb. Similarly, for a six-active-class classification of covertly spoken words, 79.17 ± 14.30% accuracy was achieved with O2Hb and 86.81 ± 9.90% with HHb as an absorber. These results indicate that a control paradigm based on covert speech can be reliably implemented into future Brain⁻Computer Interfaces (BCIs) based on NIRS.

Multiday EMG-Based Classification of Hand Motions with Deep Learning Techniques.

Pattern recognition of electromyography (EMG) signals can potentially improve the performance of myoelectric control for upper limb prostheses with respect to current clinical approaches based on direct control. However, the choice of features for classification is challenging and impacts long-term performance. Here, we propose the use of EMG raw signals as direct inputs to deep networks with intrinsic feature extraction capabilities recorded over multiple days. Seven able-bodied subjects performed six active motions (plus rest), and EMG signals were recorded for 15 consecutive days with two sessions per day using the MYO armband (MYB, a wearable EMG sensor). The classification was performed by a convolutional neural network (CNN) with raw bipolar EMG samples as the inputs, and the performance was compared with linear discriminant analysis (LDA) and stacked sparse autoencoders with features (SSAE-f) and raw samples (SSAE-r) as inputs. CNN outperformed (lower classification error) both LDA and SSAE-r in the within-session, between sessions on same day, between the pair of days, and leave-out one-day evaluation (p < 0.001) analyses. However, no significant difference was found between CNN and SSAE-f. These results demonstrated that CNN significantly improved performance and increased robustness over time compared with standard LDA with associated handcrafted features. This data-driven features extraction approach may overcome the problem of the feature calibration and selection in myoelectric control.

The effect of time on EMG classification of hand motions in able-bodied and transradial amputees.

While several studies have demonstrated the short-term performance of pattern recognition systems, long-term investigations are very limited. In this study, we investigated changes in classification performance over time. Ten able-bodied individuals and six amputees took part in this study. EMG signals were recorded concurrently from surface and intramuscular electrodes, with intramuscular electrodes kept in the muscles for seven days. Seven hand motions were evaluated daily using linear discriminant analysis and the classification error quantified within (WCE) and between (BCE) days. BCE was computed for all possible combinations between the days. For all subjects, surface sEMG (7.2 ±â€¯7.6%), iEMG (11.9 ±â€¯9.1%) and cEMG (4.6 ±â€¯4.8%) were significantly different (P < 0.001) from each other. A regression between WCE and days (1-7) was on average not significant implying that performance may be considered similar within each day. Regression between BCE and time difference (Df) in days was significant. The slope between BCE and Df (0-6) was significantly different from zero for sEMG (R2 = 89%) and iEMG (R2 = 95%) in amputees. Results indicate that performance continuously degrades as the time difference between training and testing day increases. Furthermore, for iEMG, performance in amputees was directly proportional to the size of the residual limb.

Psychophysical Evaluation of Subdermal Electrical Stimulation in Relation to Prosthesis Sensory Feedback.

This paper evaluated the psychophysical properties of subdermal electrical stimulation to investigate its feasibility in providing sensory feedback for limb prostheses. The detection threshold (DT), pain threshold (PT), just noticeable difference (JND), as well as the elicited sensation quality, comfort, intensity, and location were assessed in 16 healthy volunteers during stimulation of the ventral and dorsal forearm with subdermal electrodes. Moreover, the results were compared with those obtained from transcutaneous electrical stimulation. Despite a lower DT and PT, subdermal stimulation attained a greater relative dynamic range (i.e., PT/DT) and significantly smaller JNDs for stimulation amplitude. Muscle twitches and movements were more commonly elicited by surface stimulation, especially at the higher stimulation frequencies, whereas the pinprick sensation was more often reported with subdermal stimulation. Less comfort was perceived in subdermal stimulation of the ventral forearm at the highest tested stimulation frequency of 100 Hz. In summary, subdermal electrical stimulation was demonstrated to be able to produce similar sensation quality as transcutaneous stimulation and outperformed the latter in terms of energy efficiency and sensitivity. These results suggest that stimulation through implantable subdermal electrodes may lead to an efficient and compact sensory feedback system for substituting the lost sense in amputees.