DIY EMG Servo Control: MyoWare Muscle Sensor with Arduino UNO

While NeuroSwitch currently works with one channel, using multiple muscle groups for input provides synergistic effects rather than just additive effects. Multiple inputs allow for combinations (permutations) that enable much more complex control. A four-channel unit is already being coded for future development. The technology enables patients to control robotic arms through computer interfaces. A hobby robotic arm (available for about $70) can perform various functions from scratching a nose to passing food or moving objects on a tabletop. This demonstrates how one EMG input can reach out, talk to people, email, go on the internet, and physically interact with the environment. The technology opens up opportunities for people with disabilities to use their skills and knowledge bases without being limited by physical disabilities.

One approach to improving prosthetic control is using multiple EMG electrodes (4, 6, 8, up to 64 channels) to cover the entire forearm. This allows recognition of individual finger movements. However, this approach faces challenges with generalization, as different algorithms may be needed for each user.

Multiple servo motors can be controlled simultaneously using a single Arduino board by connecting each motor to separate pins. Each motor can be controlled independently using its own potentiometer or shared by a single potentiometer. This configuration creates multiple degrees of freedom in a robotic arm, where different joints can move at different angles simultaneously. The Arduino processes input from sensors and sends appropriate signals to each motor pin.

To control multiple servos simultaneously, modify the loop to iterate through a range of channels rather than individual servos. For example, controlling 8 servos at once involves looping from 0 to 7 on one board and 8 to 15 on another board. This approach ensures all selected servos move in unison with the same angle step. The delay between steps determines movement speed. This technique is useful for creating coordinated movements where groups of servos need to move together.

When controlling multiple servos simultaneously, the Moteus library requires creating separate controller instances for each servo, each with its own unique ID. These controllers are stored in a vector of shared pointers. The common configuration options (resolutions, query modes) are shared between controllers, while individual settings like servo ID are configured separately. Before sending commands, all controllers should be stopped to clear any faults. This approach enables coordinated control of multiple degrees of freedom in a robotic system.

ESP-NOW is a wireless communication protocol available on ESP microcontrollers that enables direct device-to-device communication without requiring a gateway, offering advantages over WiFi (no router needed, faster response time, better range) and Bluetooth (longer range), while requiring no extra hardware and supporting multiple devices through MAC address addressing; however, it has limitations including a maximum packet size of 250 bytes, no automatic packet retransmission on loss, and lower audio quality for voice applications due to the ESP32's ADC/DAC limitations.

ESP-NOW is a wireless communication protocol defined by Espressif that operates at the data link layer, reducing the traditional five-layer OSI model to a single layer for faster response times and lower resource consumption. This protocol enables direct device-to-device communication without requiring Wi-Fi connections or third-party devices, supporting one-to-many and many-to-many device connections. Key features include RSSI-based distance authorization for secure pairing, millisecond-level response speeds, and the ability to function independently as an auxiliary protocol for device provisioning, debugging, and firmware upgrades. ESP-NOW can upgrade multiple devices simultaneously in under one minute and supports large-scale installations with hundreds of devices through unicast, broadcast, and group control mechanisms.

ESP-NOW is a communication protocol developed by Espressif that enables direct device-to-device communication between multiple ESP32 boards without Wi-Fi, routers, or pairing processes. It offers extremely low power consumption because radios only wake up for fractions of a second to send each packet. A demonstration shows one ESP32 reading analog values from a potentiometer and sending them to another ESP32, which adjusts LED brightness based on received values.

ESP-NOW is a protocol developed by Expressif that enables ESP32 and ESP8266 boards to communicate directly with each other in a peer-to-peer network without requiring Wi-Fi or a router, using the 2.4GHz radio transceiver; it supports unidirectional and bidirectional communication with a maximum packet size of 250 bytes, making it ideal for applications like remote temperature sensing where multiple sensors can transmit data to a central receiver.

ESP-NOW is a low-level wireless communication protocol developed by Espressif that enables direct device-to-device communication without requiring a Wi-Fi router or Bluetooth connection; it operates at the data link and physical layers of the network stack, utilizing vendor-specific action frames within the 802.11 Wi-Fi standard to transmit up to 250 bytes of data per packet using MAC addresses for addressing, making it ideal for simple smart home applications like sensor data transmission and actuator control where reliability guarantees are not required.

TinyML (Tiny Machine Learning) enables running optimized machine learning models on low-power microcontrollers like the Wio Terminal (ARM Cortex M4F core at 120 MHz), allowing intelligent devices to perform tasks such as gesture recognition using a single light sensor. The process involves collecting sensor data with Data Forwarder, processing it through pre-processing blocks (flatten, spectral analysis, or raw data), training a neural network classifier, and deploying the model to the device for real-time inference. The video demonstrates achieving 92.4% accuracy in rock-paper-scissors gesture classification using a 35KB neural network, illustrating how TinyML can transform simple sensors into intelligent devices.

This segment covers the complete ML workflow for deploying gesture recognition on microcontrollers. It explains how ML differs from traditional programming by learning patterns from examples rather than explicit instructions. The classification approach groups similar motion data patterns into categories. The pipeline includes wireless data collection using ESP8266 flash memory, storing data in CSV files, training a random forest classifier in Python, and deploying the trained model on the microcontroller. The final system achieves approximately 95% accuracy in recognizing gestures like left/right swipes and slams.

TinyML enables deploying machine learning models on microcontrollers like Arduino boards, which have severe resource constraints (e.g., 1MB flash, 256KB RAM). The workflow involves collecting training data from sensors, training models in Python using standard Keras, converting models to TensorFlow Lite format, and converting them to C arrays for embedding in microcontroller firmware. This enables real-time inference on tiny devices with no internet connectivity.

Muscle gesture recognition devices use machine learning classifiers trained on thousands of user data points to recognize muscle patterns without requiring extensive initial training. The calibration process involves users performing preset gestures (spreading fingers, waving, fist, pinky-to-thumb) to teach the device their unique muscle anatomy and physiology. The pinky-to-thumb gesture serves as a universal unlock mechanism to prevent accidental activation during normal activities like presentations or handling objects. Applications include music/video control (wave right/left), virtual knob manipulation (hand rotation), and presentation slide navigation. The system automatically adapts to individual muscle patterns, enabling accurate gesture recognition across different users without per-person training procedures.

Tiny ML experiments can be implemented as websites using JavaScript and Web Bluetooth API, enabling anyone with a modern browser (Chrome or Edge) to interact with microcontroller devices. Practical demonstrations include playing virtual drums based on accelerometer gestures and controlling web browsers using finger movements. The system uses a two-way API where websites can send trained models to microcontrollers and receive real-time sensor data. This approach eliminates complex USB driver requirements and makes tiny ML accessible to beginners through familiar web technologies.