Technical Features and Functionalities of Myo Armband: An Overview on Related Literature and Advanced Applications of Myoelectric Armbands Mainly Focused on Arm Prostheses

The following section is reported an overview of research studies focused on possible advanced applications of the MYO armband and on related algorithms for improving the user interaction with the connected devices.

The increased integration between technology and humans, besides the studies of flexible and bendable electronic equipment such as roll up displays and wearable devices, has attracted the attention of many researchers and companies in developing even more sophisticated and complex devices with high performances, but simple to use for the users.

In particular, the employment of MYO armband, as demonstrated by numerous research studies (Sathiyanarayanan et al., 2015; Shin et al., 2015; Ganiev et al., 2016; Morais et al., 2016), is not limited to application in the prosthetic field; indeed, MYO armband can be used also in many other applications such as, for example, the development of a robotic arm piloted through natural hand movements or for controlling a mechanical device by means of hand gesture, thus avoiding the use of joysticks or other radio control devices.

Therefore, by exploiting the electrical signals generated by the contractions of the forearm/arm muscles (detected by the eight EMG electrodes) in order to control the movements of wrist, hand, and fingers (as shown in Fig. 1), MYO armband is able to drive the connected devices through the Bluetooth technology.

An application field where MYO armband is widely used is that of games; “Fruit Ninja” is a good example of how the armband can be used to slice the fruits passing on the screen using the user gestures; another example is “Race the Sun”, a game in which the user has to control the path toward the sun of a volatile in order to avoid its death (as depicted in Fig. 2).

In Figure 3, two users wearing the Myo armband have to paint the whiteboard with two different colors; the player who is able to fill the board faster (in the figure the blue one) is the winner. In Figure 4A the player, by using the Myo armband, is controlling a weapon, while in Figure 4B he is driving a kart.

Myo armband can be used to play with many other games, as reported on the Myo device website (Myo Armband web site; Myo Market, 2013–2016). Another application field of the Myo armband is related to entertainment: as an example, the famous DJ Armin Van Buuren (electronic/dance music), made Myo Armband an important part of his shows, by relating plays of lights to his movements during on-stage performances (Fig. 5).

Anyway, apart from the applications previously reported, as shown in Figure 6, Myo armband can be also used for controlling flying drones, robot movements (particularly relevant during military maneuvers, where the trained military personnel can remotely drive a robot simply by wearing and using the armband) and so on.

Furthermore, as shown in Figure 7, it is also possible to use the armband to navigate on PC desktop between the different windows, selecting and exploring several applications and performing many different tasks; a good example is its use as presentations remote controller: the user, by wearing the Myo armband, can control the presentation software with proper gestures and motion (respectively through EMG and IMU readings).

The application fields of Myo armband, as cited above, are numerous: apart from games, robotic and computer applications, it also finds application in the healthcare field. In particular the company TedCas, a leader in medical imaging, by using the Myo armband, various cameras and voice recognition software has realized a technology addressed to surgeons that during surgeries can control in real-time several medical instruments, as shown in Figure 8.

This way, the surgeons can perform by themselves many actions that should be otherwise performed by other people, keeping their hands clean and free during surgeries and thus reducing surgeries time and improving the patients’ safety by reducing the risk of infection. For example, through the armband, the surgeons can rotate a 3D image of the examined body part by simply moving the hand, or pilot a camera to visualize in detail the tissues under analysis during a surgery without any physical contact (ADORA-MED d.o.o.).

In order to illustrate the huge potential offered by the Myo armband in prosthetic applications, in the following the last progresses in this field, are reported. The Myo armband, in fact, can also be used to control the movements of a prosthesis that replaces an amputated limb. Figure 9 shows a patient with a prosthetic arm directly connected to his skeleton (osseo-integration) during an experimental test at the Johns Hopkins University (Thalmic Labs Inc., 2013–2017); the used prosthesis employs two Myo armbands on the upper arm to detect the electrical activity of the biceps and triceps muscles.

So, the patient is able to correctly grab, lift, move and release the grasped object thanks to the Myo armbands which allow precise and correct prosthesis’s movements.

Figure 10 shows the amputee patient wearing the Myo armband and a bland is used to better tighten the armband around the arm.

By moving the arm, the amputee is able to move the robotic prosthesis, attached to a support, placed at a certain distance from her. In Figure 11 the patient is raising her arm and the robotic prosthesis follows her movements.

In Figure 12, it is shown the arm wearing the Myo armband, covered by the bland; in Figure 12A the muscles contraction can be noted, while in Figure 12B, the arm muscles are relaxed. In addition, by performing proper muscles’ contractions or movements, it is possible to move even a single finger of the prosthesis, as shown in Figures 13A and B where, as an example, the prosthesis thumb is, respectively, opened and then closed.

By providing the prosthetic hand with sensors such as force sensors, temperature sensors, and contact sensors, as shown in Figure 14, the patient obtains information related to the grasped objects.

In particular, the contact sensors allow the patient to feel the consistency of the grasped objects; through some sensors/electrodes positioned on the skin as shown in Figure 15, this information is received by the arm and exploited to communicate when the prosthetic hand touches an object. In the experimental test of Figure 15, the doctor touches the prosthetic fingers and the patient receives a feedback of the touch, “feeling” the contact.

Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Fig. 8

Fig. 9

Fig. 10

Fig. 11

Fig. 12

Fig. 13

Fig. 14

Fig. 15

Nowadays several typologies of active upper-limb prostheses are available on the market: the main distinction can be made between passive prostheses, body-powered prostheses and myoelectric prostheses.

Passive upper-limb prostheses are used for cosmetic restoration and have limited if any functional capabilities. They do not generally have active prehension, so do not restore the hand functionality, but some fingers may be designed to be placed manually to improve the prosthesis function (as an example, thumb opposition). Passive prostheses are compact, light, and need little maintenance, above all regarding the cosmetic silicone glove, which is much less stressed compared to active prostheses’ one.

Body-powered prostheses are maneuvered and powered by excursion and force from gross body movements with the use of a harness system; in particular, depending on the level of amputation, the prosthetic hand is controlled either by abduction of the shoulder or flexion of the wrist. Two main typologies of body-powered hands are available: voluntary opening (VO), which are normally closed and are opened when a movement is performed, or voluntary closing (VC), which in contrast are normally opened and are closed following a shoulder/wrist movement. These hands are not expensive, robust, and light, but the functional range of motion can limit hand function. A body-powered hook is shown in Figure 16A, while in Figure 16B is shown the Child CAPP Hand prosthesis sold by the Fillauer LLC company (Fillauer LLC, 2018); the latter prosthesis, used by children, is particularly small and lightweight (just 161 g) and it is provided with an inner spring whose tension can be manually adjusted.

Myoelectric prostheses are powered by an external battery system and as explained above are operated through EMG signals derived from muscle contractions. Among myoelectric prostheses, two main device categories can be distinguished (Fig. 17): the 1 DOF (degree of freedom) prostheses are actuated by a single motor and allow one to perform the contemporary opening/closing of three fingers (thumb, forefinger and middle finger); these prostheses are light, compact, low cost and easy-to-use, but are also highly limited in the movements that the user can perform; poli-articulated prostheses, instead, employ five/six motors which allow independent movements of each finger, and thus are heavier, bulkier, and more difficult to control, but at the same time they are highly dexterous.

The main issue of poli-articulated prostheses is related to the difficulty encountered by user to control them; in fact, these devices need a greater amount of data related to the user muscular activity, compared to the 1-DOF prostheses, due to the greater number of movements that they can perform. Therefore, a more accurate processing of the EMG signals is needed besides of a larger number of muscular groups to be monitored in order to obtain an effective control of the device. These factors inevitably involve a more complex electronic system for the prosthesis management, which consequently causes an increase of the prosthesis cost, besides the higher costs associated with more complex mechanical components. The high costs of these prostheses can lead the patient to become discouraged and to decide to not wear/buy this kind of prostheses.

Figure 18 shows a poli-articulated prosthesis, i-limb hand, produced by Touch bionics (Touch Bionics Inc, 2018), with the indication of its principal components.

The bebionic hand, an advanced poli-articulated prosthesis provided by Ottobock company, provides 14 different grip patterns. The prosthesis, shown in Figure 19, is provided with five motors, one for each finger, microprocessors that continuously monitor the position of each finger, a proportional speed control to give precision control over delicate tasks, four wrist options, selectable thumb positions and auto-grip, i.e., the prosthesis automatically senses when a gripped item is slipping and adjusts the grip to secure it (Ottobock HealthCare GmbH (a)).

The hand prosthesis employs myoelectric sensors to detect the muscle’s activities and allows one to perform fingers movements in a very precise way, as shown in Figure 20.

Figure 21 shows the bebionic hand socket and the used myo-electrodes, models 13E200 = 50/13E200 = 60 or 13E202 = 50/13E202 = 60, installed within it; as shown in the schematic view of Figure 22, two myo-electrodes are used; they send, in wired mode, the detected signals to the microprocessors integrated into the prosthesis (Ottobock HealthCare GmbH (b)).

With reference to new 3D-printed devices, both the Brunel Hand by Open Bionics (shown in Figs. 23A and 23B) and the Victoria Hand prostheses (Fig. 23C) are relevant. However, these products, although available at low cost since realized by means of 3D printing technology, present limitations concerning the movements actuation. In fact, the fingers of Open Bionics hand are actuated by means of nylon tendons, which are characterized by high frictions, and by five different servo-motors, creating a bulky and heavy structure (Open bionics web site; Victoria Hand project). Another model of open source robotic hand is the InMoov Hand, shown in Figure 24, in which the fingers’ movements are actuated by employing five different servomotors positioned in the forearm (InMoov open source 3D printed life-size robot).

The Alfred Mann Foundation has created Implantable Myoelectric Sensors (IMES); these are multiple single-channel implanted EMG sensors which allow an amputee to control his/her artificial limb by means of the arm’s muscular activity. Figure 25 shows the IME sensors implanted into the arm of a patient. EMG signals, generated by the residual muscles at each implant site, are amplified and digitized by the IMES; an extra-corporeal Telemetry Controller (TC) within the limb prosthesis controls a time division multiplexing (TDM) sequence to provide power and manage RF transmissions from each implant over a common inductive link. Each IMES is wirelessly powered and telemeters an EMG signal (Troyk et al., 2007; Baker et al., 2010; Laboratory of Neural Prosthetic Research © 2018, 2018). The TC decodes the received EMG signals from all of the IMES devices, and passes the multi-channel EMG data to a Prosthesis Controller. By locating the IMES in separately innervated muscles, each IMES can be treated as an independent control site with minimal cross-talk or interference (Merrill et al., 2011; Tan et al., 2014; Laboratory of Neural Prosthetic Research © 2018;). The user, by thinking to open/close or rotate the prosthetic hand, or to move specific fingers, is able to perform many common movements, as shown in Figure 26.

However, although advancements in prosthetic limbs have been beneficial and nowadays allow amputees to carry out daily activities with reasonable ease, they do not provide the user with sensory feedback. In this context, the recent technology progresses in the prosthetic field have allowed to obtain the first upper-limb amputee in the world who is able to “feel” the grasped object in real-time with a sensory-enhanced artificial hand. The prosthetic system has been created by Silvestro Micera and colleagues from the École polytechnique fédérale de Lausanne (EPFL) in Switzerland and the Sant'Anna School of Advanced Studies (SSSA) in Italy. The prosthetic hand, containing sensors and electrodes, is wired to the user forearm and is controlled through the surface myographic electrodes integrated into the prosthesis.

When the patient touches an object, refined electrical impulses are sent to the central nervous system through other electrodes, surgically implanted into the nerves of the upper arm, allowing the patient to instantly feel it. A prototype of this system, worn by its user, is shown in Figure 27.

In order to create sensors that could detect information from touch, the scientists measured tension in artificial tendons that control finger movements. This measurement is then turned into an electrical current; however, since the central nervous system is unable to understand this kind of signals, proper developed computer algorithms are used to transform these signals into a refined impulse that can be understood by sensory nerves, as shown in Figure 28.

These new impulses are then sent through wires into four electrodes that have been surgically implanted into the nerves of the upper arm, therefore producing the sense of touch (Tan et al., 2014; Raspopovic et al., 2015).

This way, the patient is able to detect her grip strength on the objects she grasps, as well as their shape and consistency. Furthermore, recent technological progresses in prosthesis manufacturing allow one to replicate nearly perfect size and weight of the natural hand.

The main disadvantages of this prosthetic system are its high cost, related to the high level of technology implemented, and its intrusiveness (i.e., surgical operations are needed to implant the myoelectric sensors).

Two other innovative techniques for prostheses control can be mentioned: the myokinetic control interface, which aims to track the muscles contractions by means of magnetic field sensors, implanting permanent magnets into the user forearm (Tarantino et al., 2017), and the ultrasonic technology, according to which ultrasound signals can detect continuous and simultaneous movements of the arm muscles, which thanks to the use of machine learning can be associated in real time to each single finger, together with the force level the user intends to use (Georgia Institute of Technology, 2017).

With reference to commercial hand prostheses, algorithms related to the correct acquisition of the EMG signals and consequently to the correct actuation of the mechanical parts are not available on the product web-page. Anyway, for a simple prosthesis such as a hook prosthesis, the algorithms, if implemented, are very simple; in contrast, algorithms related to articulated prostheses (those featured by high dexterity) are much more complexes.

The prosthesis realized by BionIT Labs Company in collaboration with the department of Innovation Engineering of University of Salento, allows one to overcome the issue previously discussed and highlighted in red in Figure 17; the innovative underactuated mechanism on which the prosthesis named Adam’s Hand is based can actuate 15 degrees of freedom with just one motor, instead of the five–six motors conventionally used in other prosthetic devices. The torque is automatically distributed among the fingers that adapt to the specific form of the grasped object, always developing the most stable grip. The transmission is based on geared differential modules, stiff and compact, that exert constant forces independent of the fingers kinematic state. The fingers can also move independently through the use of dedicated clutches or brakes (Zappatore et al., 2016; Zappatore et al., 2017). This allows one to simplify the control logic and to save on prosthesis cost, weight, and dimensions, together with a lower power consumption and a less noisy actuation linked to the use of a single motor (as indicated in Fig. 29).

The realized prosthesis is also designed in a modular way, in order to adapt to different amputation levels: the main unit is provided with artificial intelligence and can also be used for post-traumatic rehabilitation procedures and muscular training for non-amputee patients, while the hand module can be also mounted on robotic arms in order to assist, eventually through EEG, completely paralyzed users.

Thanks to Myo armband, user gestures can be recognized and an algorithm, that uses rectified and filtered EMG and 9-axes IMU signals, evaluates the strength required to actuate each movement. These data are time-averaged and given as input to a single layer neural network that is initialized through a training phase performed at the system start.

The myoelectric prosthesis is also provided with sensors and actuators: the LM35 temperature and FSR-400 pressure sensors, installed into each fingertip, communicate the acquired data to the Arduino Micro board on which the developed custom board is based and, by means of the Raspberry Pi board housed in the prosthesis socket, these data are visualized on the touchscreen display in order to provide a feedback to the user (Primiceri et al., 2016; Visconti et al., 2016; Visconti et al., 2017a, 2017b;).

The Arduino micro board receives the Myo armband EMG and IMU data through a HM-11 BLE module, while the recognized poses are reported on the touchscreen display in real-time. The realized electronic control board, the switching regulator used for the Raspberry Pi board power supply and the touch-screen mounted on the Raspberry Pi board, together with the related housing into the prosthesis socket, are shown in Figure 30A; the employed Lithium battery and the Myo armband are also indicated (Patrono et al., 2017; Visconti et al. 2017c, 2017d).

The employed mechanism is based on the use of a single DC motor for fingers movement and two servomotors for wrist movement, unlike traditional poli-articulated prostheses, which employ from five to six motors to drive and control each finger motion, as discussed above.

The use of a single motor to actuate the whole hand allows one to simplify the prosthesis control, as well as to save on its cost and weight, also improving the device autonomy. Moreover, data concerning EMG and IMU signals are processed by a properly developed Wolfram Mathematica software and can be monitored remotely by the orthopedic staff that follows the patient during the rehabilitation period (Fig. 30B), thanks to the Wi-Fi connectivity provided by the Raspberry PI board.

Fig. 16

Fig. 17

Fig. 18

Fig. 19

Fig. 20

Fig. 21

Fig. 22

Fig. 23

Fig. 24

Fig. 25

Fig. 26

Fig. 27

Fig. 28

Fig. 29

Fig. 30

The Myo armband, shown in Figure 31, is produced by the Thalmic Labs Canadian company and, as discussed above, it is mainly used to acquire the myoelectric signals produced by the user forearm muscles (Myo Armband web site; Thalmic Labs Inc., 2013–2018).

Fig. 31

The device is provided with eight EMG electrodes, a 9-axes IMU composed of a 3-axes accelerometer, a 3-axes gyroscope and a 3-axes magnetometer and a vibration motor used to alert the user when a particular event occurs. Its technical specifications are reported below:

Freescale Kinetis ARM Cortex M4 120Mhz MK22FN1M MCU;

The armband operation is managed by ARM Cortex M4 processor, while data transmission are performed through the BLE NRF51822 chip, which exchanges data with HM-11 BLE module mounted on bottom side of the prosthesis driving/control unit. In Figure 32A, MYO armband is shown disassembled while Figure 32B shows the electronic control board embedded into the main central element with the previously described sections highlighted in different colors.

In particular, the BLE NRF51822 chip for the wireless transmission is highlighted in blue, the ARM Cortex M4 microcontroller in red, the antenna which transmits the data in grey, the vibration motor for the user feedback in brown and the micro-USB connector in purple. The latter is used to recharge the battery and to update the device firmware.

Two rechargeable lithium batteries, shown in Figure 33, are used to supply the armband; these are located into two elements of the armband and can be recharged providing a voltage value of 5 V by means of the micro-USB connector.

EMG Electrodes employed for muscles activity detection

The correct detection of the EMG signals, by means of the employed EMG electrodes, is a key factor to consider in order to properly control the prosthesis. In order to acquire the EMG signals, the voltage value/electric field associated with the muscles activity has to be detected; this potential is produced through the electrical depolarization of the muscular fibers in response to an electric pulse provided from the nervous system. Bipolar electrodes are used for the measurement of the muscular signals, which will be then amplified by using proper amplification circuits (Carlo, 1997), (Criswell, 1998). Two typologies of EMG sensors can be used in this kind of application: the insertion sensors and the surface sensors.

The EMG insertion sensors (shown in Fig. 34A) allow obtaining very accurate and localized measurements of the muscles activity; the disadvantage of this type of sensors is that they are very invasive since they have to be inserted into the user’s muscle through the use of a needle. The use of this typology of EMG sensors in the context of the realized prosthesis was excluded; in fact, Adam’s Hand does not require so detailed information on a specific muscle area and, in addition, the use of needles could be impractical for the patient and even cause infections. This is the reason they are not used on commercial prostheses.

On the other hand, the EMG surface electrodes employed by Myo armband can provide information about activity of the forearm muscles shown in Figure 34B; this information, even if not detailed, is still sufficient to control the realized prosthesis, due to its ease of control; moreover, they are not invasive and are easy to integrate into the prosthesis socket.

The EMG surface electrodes are divided into two main categories: the passive and the active (differential) ones. The formers are composed of a single conductor and needs a conductive material which guarantee an optimal coupling with the skin. The active electrodes, instead, are composed of three conductors; the central one is located close to the analyzed muscle and detects the muscular activity, while the other two, placed before and after the central one along the same muscle, act as filters and provide noise immunity, allowing for a better signal reading. In particular, as shown in Figure 35A, Myo armband employs eight active EMG electrodes located in eight different areas of the forearm; the ST78589 operational amplifier, shown in Figures 35B and 35C, is used for the amplification of the signal provided from the EMG electrodes and it is integrated into each element of the armband.

Fig. 32

Fig. 33

Fig. 34

Fig. 35

Fig. 36

Fig. 37

In this research work, the human–machine interaction performed by using the last generation device Myo armband is analyzed and discussed, with particular focus to healthcare applications such as upper-limb prostheses. The Myo armband can be used in numerous applications due to its excellent technical features and ease of use; it is provided with eight electromyography electrodes, a 9-axes IMU, and a transmission module; all the acquired data are sent wirelessly (through BLE technology) to the control electronics to use actuators or perform other specific tasks. After an overview on Myo armband application fields and on the latest research works related to its use in prosthetic applications, we analyzed and discussed in detail its technical features and functionalities.

A trans-radial myoelectric prosthesis prototype, named Adam’s Hand and realized by BionIT Labs company in collaboration with our research department is also presented; it is equipped with sensors and actuators that move the hand and the wrist and which are controlled by the user myoelectric signals detected through Myo armband: the related EMG data, together with IMU data, are in fact sent wirelessly to the control electronics in order to activate a DC motor that handles the five fingers and two servomotors that actuate the wrist. The realized driving/control electronic unit manages the different prosthesis modules: it acquires data from five force sensitive resistors (FSR) and five LM35 temperature sensors, one for each finger, it drives two Futaba S3305 servo-motors used for wrist movement and a Maxon DCX 19S DC motor used for fingers movement. The Arduino-based control unit receives data from Myo armband by using the HM11 module that integrates SoC Texas Instruments CC2541 chip and it exchanges data with a Raspberry Pi board (placed into a properly realized case which also houses a LCD display). The Raspberry Pi board collects these data and, through Wi-Fi connection, sends them on cloud to a dedicated web server to be monitored remotely by the orthopedic staff. Further Adam’s Hand details will be described in the future; anyway, the realized prosthesis confirms that Myo armband is an optimum candidate to be used in these kind of applications, allowing to get reliable, robust and low cost electromyographic data.