North (formerly Thalmic Labs), the creator of the Myo armband, was acquired by Google in June 2020. Myo sales ended in October 2018 and Myo software, hardware and SDKs are no longer available or supported. Learn more.


Reliability of Gesture Recognition Systems

Reliability of Gesture Recognition Systems

Gesture recognition is a new controlling scheme that can be developed to communicate with electronic devices including phones, computers, smart TVs, and even new wearable devices such as Google Glass and the Oculus Rift. Several gesture recognition schemes have been introduced in recent years including wired gloves and camera-based detection. Here at Thalmic Labs, we’re working on a completely new gesture recognition scheme for the Myo™ armband: an electromyography (EMG)-based system.

The Myo armband is an all-in-one gesture recognition system that is capable of recognizing different sets of hand gestures based on electrical signals from the forearm’s muscles. These signals are captured by the advanced sensors that we’ve designed in-house and processed by our embedded algorithms. The gesture recognition result is output over Bluetooth 4.0 Low Energy to connect with digital devices.

To achieve reliable gesture recognition, the machine learning team deals with several challenges such as inter- and intra-variability in performing gestures. Inter- and intra-variability refer to the differences in the way that multiple people perform the same gesture and the way that a single individual performs the same gesture each time, respectively. These differences in gestures could include the style and the pace of performing various gestures.

The intra-variability issue is less significant than inter-variability, since a person performs a gesture similarly most of the time and a small variation in each trial can be easily compensated by the artificial intelligence algorithm laid out in the system.

While inter-variability is not usually found in simple hand gestures, more complex gestures require the use of population models to map out gestures. Having a large number of individuals in our data-set, we can map out gestures to detect what the user is intending to do when a given gesture is made. The goal is to properly represent similarities and differences among individuals and trials, which is no trivial task.

According to the specific application, the number of gestures that must be automatically recognized by the Myo device can vary. New gesture recognition systems like the Myo armband may not provide the same number of distinct inputs as the keyboard, but can combine motions and gestures to achieve powerful interactions. Before Myo, it was nearly impossible to imagine having free hands while remotely controlling a device from far away. Wearable devices like the Myo armband are finally making this dream a reality!

You've successfully subscribed to The Lab!