Muscle Activity Monitoring with Arduino EMG Sensor | Electromyography Tutorial

Bilateral upper limb amputation often results from electrical accidents. Traditional prosthetics use hooks and wires transferring chest muscle movements via cables. Robotic limbs with 18 motors and 18 degrees of freedom offer near-lifelike replacement. Myoelectric (EMG) signals are electrical potentials muscles produce as they receive nerve impulses. Residual muscles in amputees can produce signals, but surface electrodes cannot easily isolate individual signals due to muscle group mixing. Implantable myoelectric sensors placed inside individual muscles could provide multiple sensing sites for enhanced prosthetic control. These devices are small (penny-sized) with ceramic tubes, electrodes, coils for wireless power, and integrated circuits for signal processing. Testing in rhesus monkeys showed signals recorded from normal arms played back to robotic arms, demonstrating untethered devices that float with muscles to pick up signals wirelessly.

Myoelectric robotic prosthetics are devices that use electrodes to detect electrical signals from residual muscles, translating muscle contractions into controlled movements of the prosthetic limb; for example, wrist flexion signals trigger the prosthetic hand to close while wrist extension signals trigger it to open, allowing users to perform complex tasks like gripping, pointing, or using a computer mouse, though these devices are heavier, more expensive, and less durable than conventional prosthetics, making them suitable only for specific patients and activities.

Myoelectric prosthetics consist of socket, electrodes, battery, and motor components. Electrodes detect muscle signals and convert them to mechanical movements. Channel systems (2, 4, 6) determine control complexity. 2-channel systems control basic hand functions. 4-channel systems add wrist rotation. 6-channel systems include elbow control. Grip patterns are controlled through specific muscle contractions. Mode switching allows patients to change between functions. Pediatric prosthetics require special considerations for weight, durability, and growth. Advanced features include multiple grip patterns and sensors. Training involves learning to generate consistent muscle activation patterns for reliable prosthetic control.

Myoelectric prosthetics are artificial devices that substitute for missing body structures by using electrical signals from muscle contractions (captured through electromyography) to control movement, but they face significant challenges including the complexity of human hand biomechanics (27 degrees of freedom), the variability of myoelectric signals between individuals, and the difficulty of achieving natural movement control and tactile feedback.

In 1948, Reinhard Raiter created the first myoelectric prosthetic using vacuum tube amplifiers to detect muscle signals and control motors. Twenty years later, Soviet engineer Alexander Kabrinsky solved the miniaturization problem using transistors, creating a wearable device. This principle—detecting muscle signals and converting them to motor movements—remains the foundation of all modern myoelectric prosthetics. Despite 30+ years of improvements, the fundamental architecture has remained unchanged. The FDA approved the i-Limb in 2014 with 10 motorized degrees of freedom, but the gap between laboratory demonstrations and real-world usability remains significant.

This extensive section presents the complete methodology and evaluation of the EMG-based hand gesture recognition system. The data acquisition system uses a Myo Armpit sensor with eight channels connected via Bluetooth, recording four gesture classes with three-second duration each. Preprocessing involves normalization, segmentation, and Empirical Mode Decomposition (EMD) to decompose signals into Intrinsic Mode Functions, selecting lower-frequency components (IMF2, IMF3, IMF4) to extract the Region of Interest. Feature extraction employs spectral domain features including roll-off point, flatness, crest, decrease, slope, and spread. Classification utilizes Support Vector Machine (SVM) with multiple kernels, achieving 98.9% accuracy with the cubic kernel. Performance evaluation uses metrics derived from confusion matrices: TP, TN, FN, FP, accuracy, precision, sensitivity, and specificity. Comparative analysis demonstrates superiority over existing approaches (98.9% vs. 98%, 85.08%, 92.98%, 96.48%). The system achieves 99% sensitivity, 98.82% specificity, and 99% precision with training time of 1.664 seconds, validating its effectiveness for prosthetic applications and assistive technologies.

Electromyography (EMG) measures electrical activity generated during muscular contractions using surface electrodes. This physiological signal can be recorded, processed, and passed to machine learning algorithms which can discriminate between different muscle contractions. This technology effectively turns the human body into a controller where muscular contractions or gestures are associated with interactive commands, enabling hands-free interaction with computing systems.

Gesture recognition is possible through the field of pattern recognition, which according to ACM's 2012 classification falls under computational methodologies. The system developed uses ensemble methods and machine learning tools. Machine learning and pattern recognition are complementary approaches that support each other, with machine learning focusing on algorithms that learn from data and pattern recognition focusing on identifying regularities in data.

In clinical settings, most myoelectric prostheses use a two-channel hardware-based control system. Two EMG sensors record signals from flexor and extensor muscle groups. Hardware-based feature extraction extracts two signatures from these signals. A simple table-based logic determines the output: if both muscles are below their respective thresholds, nothing happens; if only one muscle is activated, the hand closes or opens accordingly; if both muscles are active simultaneously, different functions like grip selection or wrist rotation occur. This system has remained remarkably stable for over a century and is widely used in clinical practice despite being less sophisticated than pattern recognition approaches.

This video demonstrates a mixed reality gaming experience where players interact with a virtual environment through gesture recognition using a Myo armband sensor, which captures accelerometer, gyroscope, orientation, and EMG data to control virtual neurons and generate music in real-time through a machine learning system.

Clinical applications of biomechanics focus on analyzing movement patterns, categorizing muscle work types, and applying principles to rehabilitation. Movement patterns can be categorized based on purpose and complexity, from basic reaching and grasping to complex locomotion. Understanding opposite directions in movement helps analyze mechanics and identify imbalances. Muscle work categorization includes concentric work (muscle shortening while generating force), eccentric work (muscle lengthening while controlling movement), and isometric work (muscle maintaining length while generating force), each with specific applications in exercise programming, rehabilitation, and athletic training. Kinematic analysis describes motion of body segments without considering forces (position, velocity, acceleration), while kinetic analysis describes forces that cause or result from motion (muscle forces, ground reaction forces, gravitational forces). Together, kinematic and kinetic analysis provides complete understanding of movement mechanics for diagnosing disorders and guiding treatment interventions. Physical activities involve cycles of loading and response, with different tissues having different loading tolerances and response characteristics.

Physical activity generates mechanical stimuli through muscle contractions that interact with the nervous system. The Golgi tendon organs and muscle spindles detect tension and changes in muscle length, sending information to the central nervous system. Human movement requires complex interactions between muscles, external forces (gravity, reaction forces), and bone articulations. Force plates measure ground reaction forces in three dimensions during movement. Normal walking shows characteristic force patterns: initial peak (20% body weight) at foot contact, decrease during mid-stance, and second peak at push-off for propulsion. Baropodometry measures plantar pressure distribution across foot regions. Abnormal patterns indicate pathology (e.g., increased rearfoot pressure, decreased midfoot pressure). Surface electromyography (EMG) measures muscle activation patterns during movement. Normal conditions show organized activation patterns (active-inactive-active-inactive), while pathological conditions show disorganized, hyperactive, or hypotonic patterns. Biofeedback allows patients to see their muscle activity in real-time, increasing motivation and adherence to rehabilitation programs. Effective rehabilitation requires determining appropriate 'doses' of mechanical stimulation: force magnitude, number of repetitions, number of sets, and frequency per week. Quality of muscle recruitment (how muscles are activated in sequence) is as important as magnitude. Biomechanical analysis helps identify injury risk factors—for example, landing forces exceeding twice maximum force indicate injury risk. Asymmetry between limbs indicates uneven loading and increased injury risk. Biomechanical assessment (force measurements, EMG, pressure analysis) provides objective data about movement function. This information, combined with clinical judgment, helps identify factors causing movement problems. The goal is to achieve symmetry in force distribution and muscle activation. Small asymmetries in muscle function can have significant impacts on performance and injury risk over time. Plantar pressure analysis can guide orthotic design, but orthotics are passive compensations. The best approach combines orthotics with active rehabilitation (flexibility, mobility, strengthening). Long-term high heel use causes structural changes: gastrocnemius shortening, increased arch height, and muscle weakness. Virtual reality environments provide challenging motor tasks that improve balance and coordination. Artificial intelligence is being integrated into rehabilitation to analyze movement data and personalize treatment. Patient awareness and understanding of exercise benefits significantly improves adherence to rehabilitation programs.

Frequency domain analysis decomposes EMG signals into constituent frequencies, revealing motor unit characteristics. Median frequency analysis detects muscle fatigue: as fatigue develops, median frequency decreases because fatigued motor units fire at lower frequencies. This creates an abrupt frequency shift at metabolic thresholds correlating with respiratory compensation and lactate accumulation. EMG research guides exercise prescription by quantifying muscle activation patterns. Different exercises preferentially activate specific muscles: unilateral bridges activate gluteus medius more than gluteus maximus; step-down exercises emphasize vastus medialis obliquus over tibialis anterior; clamshells and side planks target gluteus medius over tensor fasciae latae. These findings inform rehabilitation programming for conditions like patellofemoral pain syndrome and scapular dyskinesis. By selecting exercises that maximize desired muscle activation while minimizing unwanted co-activation, clinicians can improve rehabilitation outcomes for patients with muscle imbalances and movement disorders.

Biomechanics employs two analytical approaches: kinematics (describing movement appearance without forces—measuring range of motion in degrees) and kinetics (analyzing forces causing movement). The Oxford Muscle Strength Scale provides qualitative grading from 0-5, while dynamometers offer quantitative force measurement in Newtons. Biomechanics applies to diverse populations including athletes, elderly patients with gait/balance issues, stroke survivors, cane users, and pregnant women. It improves quality of life through optimized prostheses, footwear design, and movement analysis, demonstrating its broad clinical utility beyond sports medicine.

Gait analysis combines kinematics (movement geometry without forces) and kinetics (force evaluation including muscle forces, gravity, and ground reaction forces). Modern gait analysis uses technology: infrared motion capture cameras (200 fps), high-speed video, force plates, and surface electromyography. Reflective markers are placed on anatomical landmarks (knees, ankles, thumb-index finger joint). Software processes data to generate detailed reports for medical and rehabilitation teams. Applications include evaluating congenital disorders, neurological conditions (stroke, cerebral palsy, Parkinson's), and assessing treatment outcomes from medications, physical therapy, or orthotic interventions. The technology enables objective, real-time measurement of gait parameters including cadence, step length, step width, velocity, and joint kinematics. This allows precise diagnosis and targeted rehabilitation interventions.