Search

In recent years, deep learning algorithms have become increasingly more prominent for their unparalleled ability to automatically learn discriminant features from large amounts of data. However, within the field of electromyography-based gesture recognition, deep learning algorithms are seldom employed as they require an unreasonable amount of effort from a single person, to generate tens of thousands of examples. This paper's hypothesis is that general, informative features can be learned from the large amounts of data generated by aggregating the signals of multiple users, thus reducing the recording burden while enhancing gesture recognition. Consequently, this paper proposes applying transfer learning on aggregated data from multiple users while leveraging the capacity of deep learning algorithms to learn discriminant features from large datasets. Two datasets comprised 19 and 17 able-bodied participants, respectively (the first one is employed for pre-training), were recorded for this work, using the Myo armband. A third Myo armband dataset was taken from the NinaPro database and is comprised ten able-bodied participants. Three different deep learning networks employing three different modalities as input (raw EMG, spectrograms, and continuous wavelet transform (CWT)) are tested on the second and third dataset. The proposed transfer learning scheme is shown to systematically and significantly enhance the performance for all three networks on the two datasets, achieving an offline accuracy of 98.31% for 7 gestures over 17 participants for the CWT-based ConvNet and 68.98% for 18 gestures over 10 participants for the raw EMG-based ConvNet. Finally, a use-case study employing eight able-bodied participants suggests that real-time feedback allows users to adapt their muscle activation strategy which reduces the degradation in accuracy normally experienced over time.

This paper presents a control interface to translate the residual body motions of individuals living with severe disabilities, into control commands for body-machine interaction. A custom, wireless, wearable multi-sensor network is used to collect motion data from multiple points on the body in real-time. The solution proposed successfully leverage electromyography gesture recognition techniques for the recognition of inertial measurement units-based commands (IMU), without the need for cumbersome and noisy surface electrodes. Motion pattern recognition is performed using a computationally inexpensive classifier (Linear Discriminant Analysis) so that the solution can be deployed onto lightweight embedded platforms. Five participants (three able-bodied and two living with upper-body disabilities) presenting different motion limitations (e.g. spasms, reduced motion range) were recruited. They were asked to perform up to 9 different motion classes, including head, shoulder, finger, and foot motions, with respect to their residual functional capacities. The measured prediction performances show an average accuracy of 99.96% for able-bodied individuals and 91.66% for participants with upper-body disabilities. The recorded dataset has also been made available online to the research community. Proof of concept for the real-time use of the system is given through an assembly task replicating activities of daily living using the JACO arm from Kinova Robotics., Comment: 10 pages, submitted for review

This paper presents a wireless multichannel surface electromyography (sEMG) sensor which features a custom 0.13μm CMOS mixed-signal system-on-chip (SoC) analog frontend circuit. The proposed sensor includes 10 sEMG recording channels with tunable bandwidth (BW) and analog-to-digital converter (ADC) resolution. The SoC includes 10x bioamplifiers, 10x 3 rd order ΔΣ MASH 1-1-1 ADC, and 10x on-chip decimation filters (DF). This SoC provides the sEMG samples data through a serial peripheral interface (SPI) bus to a microcontroller unit (MCU) that then transfers the data to a wireless transceiver. We report sEMG waveforms acquired using a custom multichannel electrode module, and a comparison with a commercial grade system. Results show that the proposed integrated wireless SoC-based system compares well with the commercial grade sEMG recording system. The sensor has an input-referred noise of 2.5 μVrms (BW of 10-500 Hz), an input-dynamic range of 6 mVpp, a programmable sampling rate of 2 ksps, for sEMG, while consuming only 7.1 μW/Ch for the SoC (w/ ADC & DF) and 21.8 mW of power for the sensor (Transceiver, MCU, etc.). The system lies on a 1.5 × 2.0 cm 2 printed circuit board and weights <; 1 g.

This paper presents a new multimodal control interface for people living with upper-body disabilities based on a wearable wireless sensor network. The proposed body-machine interface is modular and can be easily adapted to the residual functional capacities (RFCs) of different users. A custom data fusion algorithm has been developed for emulating a joystick control using head motion measured with a lightweight wireless inertial sensor enclosed in a headset. The wearable network can include up to six modular sensor nodes which can be used simultaneously to read different RFCs including gesture and muscular activity, and translate them into commands. Sensor data fusion is performed inside the sensor nodes in order to free the wireless link and the base station, and decrease power consumption. Requirements of such an interface are established for people using powered-wheelchairs, and a proof of concept system is implemented and used to control an assistive robotic arm. It is shown that the performance of the system compares well to conventional control systems like the joystick controller, while being potentially more suitable for the severely disabled.

Increasing performance while decreasing the cost of sEMG prostheses is an important milestone in rehabilitation engineering. The different types of prosthetic hands that are currently available to patients worldwide can benefit from more effective and intuitive control. This paper presents a real -time approach to classify finger motions based on surface electromyography (sEMG) signals. A multichannel signal acquisition platform implemented using components of f the shelf is use d to record forearm sEMG signals from 7 channels. sEMG pattern classification is performed in real time, using a Linear Discriminant Analysis approach. Thirteen hand motions can be successfully identified with an accuracy of up to 95. 8% and of 92. 7% on average for 8 participants, with an updated prediction every 192 ms.

Assistive Technologies (ATs) also called extrinsic enablers are useful tools for people living with various disabilities. The key points when designing such useful devices not only concern their intended goal, but also the most suitable human-machine interface (HMI) that should be provided to users. This paper describes the design of a highly intuitive wireless controller for people living with upper body disabilities with a residual or complete control of their neck and their shoulders. Tested with JACO, a six-degree-of-freedom (6-DOF) assistive robotic arm with 3 flexible fingers on its end-effector, the system described in this article is made of low-cost commercial off-the-shelf components and allows a full emulation of JACO's standard controller, a 3 axis joystick with 7 user buttons. To do so, three nine-degree-of-freedom (9-DOF) inertial measurement units (IMUs) are connected to a microcontroller and help measuring the user's head and shoulders position, using a complementary filter approach. The results are then transmitted to a base-station via a 2.4-GHz low-power wireless transceiver and interpreted by the control algorithm running on a PC host. A dedicated software interface allows the user to quickly calibrate the controller, and translates the information into suitable commands for JACO. The proposed controller is thoroughly described, from the electronic design to implemented algorithms and user interfaces. Its performance and future improvements are discussed as well.

In the realm of surface electromyography (sEMG) gesture recognition, deep learning algorithms are seldom employed. This is due in part to the large quantity of data required for them to train on. Consequently, it would be prohibitively time consuming for a single user to generate a sufficient amount of data for training such algorithms. In this paper, two datasets of 18 and 17 able-bodied participants respectively are recorded using a low-cost, low-sampling rate (200Hz), 8-channel, consumer-grade, dry electrode sEMG device named Myo armband (Thalmic Labs). A convolutional neural network (CNN) is augmented using transfer learning techniques to leverage inter-user data from the first dataset and alleviate the data generation burden imposed on a single individual. The results show that the proposed classifier is robust and precise enough to guide a 6DoF robotic arm (in conjunction with orientation data) with the same speed and precision as with a joystick. Furthermore, the proposed CNN achieves an average accuracy of 97.81% on seven hand/wrist gestures on the 17 participants of the second dataset.

Increasing performance while decreasing the cost of sEMG prostheses is an important milestone in rehabilitation engineering. The different types of prosthetic hands that are currently available to patients worldwide can benefit from more effective and intuitive control. This paper presents a real -time approach to classify finger motions based on surface electromyography (sEMG) signals. A multichannel signal acquisition platform implemented using components of f the shelf is use d to record forearm sEMG signals from 7 channels. sEMG pattern classification is performed in real time, using a Linear Discriminant Analysis approach. Thirteen hand motions can be successfully identified with an accuracy of up to 95. 8% and of 92. 7% on average for 8 participants, with an updated prediction every 192 ms.

This paper presents a new multimodal control interface for people living with upper-body disabilities based on a wearable wireless sensor network. The proposed body-machine interface is modular and can be easily adapted to the residual functional capacities (RFCs) of different users. A custom data fusion algorithm has been developed for emulating a joystick control using head motion measured with a lightweight wireless inertial sensor enclosed in a headset. The wearable network can include up to six modular sensor nodes which can be used simultaneously to read different RFCs including gesture and muscular activity, and translate them into commands. Sensor data fusion is performed inside the sensor nodes in order to free the wireless link and the base station, and decrease power consumption. Requirements of such an interface are established for people using powered-wheelchairs, and a proof of concept system is implemented and used to control an assistive robotic arm. It is shown that the performance of the system compares well to conventional control systems like the joystick controller, while being potentially more suitable for the severely disabled.

Assistive technology (AT) tools and appliances are being more and more widely used and developed worldwide to improve the autonomy of people living with disabilities and ease the interaction with their environment. This paper describes an intuitive and wireless surface electromyography (sEMG) based body-machine interface for AT tools. Spinal cord injuries at C5-C8 levels affect patients' arms, forearms, hands, and fingers control. Thus, using classical AT control interfaces (keypads, joysticks, etc.) is often difficult or impossible. The proposed system reads the AT users' residual functional capacities through their sEMG activity, and converts them into appropriate commands using a threshold-based control algorithm. It has proven to be suitable as a control alternative for assistive devices and has been tested with the JACO arm, an articulated assistive device of which the vocation is to help people living with upper-body disabilities in their daily life activities. The wireless prototype, the architecture of which is based on a 3-channel sEMG measurement system and a 915-MHz wireless transceiver built around a low-power microcontroller, uses low-cost off-the-shelf commercial components. The embedded controller is compared with JACO's regular joystick-based interface, using combinations of forearm, pectoral, masseter, and trapeze muscles. The measured index of performance values is 0.88, 0.51, and 0.41 bits/s, respectively, for correlation coefficients with the Fitt's model of 0.75, 0.85, and 0.67. These results demonstrate that the proposed controller offers an attractive alternative to conventional interfaces, such as joystick devices, for upper-body disabled people using ATs such as JACO.

In recent years, deep learning algorithms have become increasingly more prominent for their unparalleled ability to automatically learn discriminant features from large amounts of data. However, within the field of electromyography-based gesture recognition, deep learning algorithms are seldom employed as they require an unreasonable amount of effort from a single person, to generate tens of thousands of examples. This work's hypothesis is that general, informative features can be learned from the large amounts of data generated by aggregating the signals of multiple users, thus reducing the recording burden while enhancing gesture recognition. Consequently, this paper proposes applying transfer learning on aggregated data from multiple users, while leveraging the capacity of deep learning algorithms to learn discriminant features from large datasets. Two datasets comprised of 19 and 17 able-bodied participants respectively (the first one is employed for pre-training) were recorded for this work, using the Myo Armband. A third Myo Armband dataset was taken from the NinaPro database and is comprised of 10 able-bodied participants. Three different deep learning networks employing three different modalities as input (raw EMG, Spectrograms and Continuous Wavelet Transform (CWT)) are tested on the second and third dataset. The proposed transfer learning scheme is shown to systematically and significantly enhance the performance for all three networks on the two datasets, achieving an offline accuracy of 98.31% for 7 gestures over 17 participants for the CWT-based ConvNet and 68.98% for 18 gestures over 10 participants for the raw EMG-based ConvNet. Finally, a use-case study employing eight able-bodied participants suggests that real-time feedback allows users to adapt their muscle activation strategy which reduces the degradation in accuracy normally experienced over time., Comment: Source code and datasets available: https://github.com/Giguelingueling/MyoArmbandDataset

In the realm of surface electromyography (sEMG) gesture recognition, deep learning algorithms are seldom employed. This is due in part to the large quantity of data required for them to train on. Consequently, it would be prohibitively time consuming for a single user to generate a sufficient amount of data for training such algorithms. In this paper, two datasets of 18 and 17 able-bodied participants respectively are recorded using a low-cost, low-sampling rate (200Hz), 8-channel, consumer-grade, dry electrode sEMG device named Myo armband (Thalmic Labs). A convolutional neural network (CNN) is augmented using transfer learning techniques to leverage inter-user data from the first dataset and alleviate the data generation burden imposed on a single individual. The results show that the proposed classifier is robust and precise enough to guide a 6DoF robotic arm (in conjunction with orientation data) with the same speed and precision as with a joystick. Furthermore, the proposed CNN achieves an average accuracy of 97.81% on seven hand/wrist gestures on the 17 participants of the second dataset.

Assistive technology (AT) tools and appliances are being more and more widely used and developed worldwide to improve the autonomy of people living with disabilities and ease the interaction with their environment. This paper describes an intuitive and wireless surface electromyography (sEMG) based body-machine interface for AT tools. Spinal cord injuries at C5-C8 levels affect patients' arms, forearms, hands, and fingers control. Thus, using classical AT control interfaces (keypads, joysticks, etc.) is often difficult or impossible. The proposed system reads the AT users' residual functional capacities through their sEMG activity, and converts them into appropriate commands using a threshold-based control algorithm. It has proven to be suitable as a control alternative for assistive devices and has been tested with the JACO arm, an articulated assistive device of which the vocation is to help people living with upper-body disabilities in their daily life activities. The wireless prototype, the architecture of which is based on a 3-channel sEMG measurement system and a 915-MHz wireless transceiver built around a low-power microcontroller, uses low-cost off-the-shelf commercial components. The embedded controller is compared with JACO's regular joystick-based interface, using combinations of forearm, pectoral, masseter, and trapeze muscles. The measured index of performance values is 0.88, 0.51, and 0.41 bits/s, respectively, for correlation coefficients with the Fitt's model of 0.75, 0.85, and 0.67. These results demonstrate that the proposed controller offers an attractive alternative to conventional interfaces, such as joystick devices, for upper-body disabled people using ATs such as JACO.

Assistive Technologies (ATs) also called extrinsic enablers are useful tools for people living with various disabilities. The key points when designing such useful devices not only concern their intended goal, but also the most suitable human-machine interface (HMI) that should be provided to users. This paper describes the design of a highly intuitive wireless controller for people living with upper body disabilities with a residual or complete control of their neck and their shoulders. Tested with JACO, a six-degree-of-freedom (6-DOF) assistive robotic arm with 3 flexible fingers on its end-effector, the system described in this article is made of low-cost commercial off-the-shelf components and allows a full emulation of JACO's standard controller, a 3 axis joystick with 7 user buttons. To do so, three nine-degree-of-freedom (9-DOF) inertial measurement units (IMUs) are connected to a microcontroller and help measuring the user's head and shoulders position, using a complementary filter approach. The results are then transmitted to a base-station via a 2.4-GHz low-power wireless transceiver and interpreted by the control algorithm running on a PC host. A dedicated software interface allows the user to quickly calibrate the controller, and translates the information into suitable commands for JACO. The proposed controller is thoroughly described, from the electronic design to implemented algorithms and user interfaces. Its performance and future improvements are discussed as well.