Gesture recognition is a powerful tool for human-computer interaction. It is increasingly redefining how we interact with our smart phones, wearable devices, televisions, and gaming consoles. QueSSence™ On Air Hand Digit Pattern Recognition is an Interactive ML application designed to make real-time machine learning and gesture recognition more accessible for non- specialists. Emphasis is placed on ease of use, with a consistent, minimalist design that promotes accessibility while supporting flexibility and customization for advanced users. Nevertheless, while a diverse range of individuals now has access to powerful sensors and rapid-prototyping tools, performing real-time gesture recognition can pose a challenge, even to accomplished developers and engineer. In addition to the increasing prevalence of gesture-based interactions in consumer devices, a diverse range of individuals are gaining access to affordable sensor technology. Many of these applications are primarily designed for off-line analysis done by domain experts and require substantial effort to recognize real-time signals. At the same time, a vast array of signal processing and machine learning techniques have been developed by researchers and industry to detect a wide range of phenomena using sensor data. Powerful libraries such as SciKit-Learn, Scipy offer many sophisticated tools, but they may not suitable for Memory and Ultra Low power constraint devices like (Redpine RS1400 chip-sets). To address this issue, and to ensure flexibility of the embedded system, we developed with a set of ML examples, motivated by techniques published in the literature and encountered as aspirational goals in our industrial experience. These motivating examples involve different sensors and classifiers, using their diverse characteristics to inform the system’s functionality and design. This Example Applications is a general-purpose tool for facilitating non specialists to create their own real-time machine-learning based systems. QueMLib performs common functionality, such as passing data between algorithms or to pre-process and feature extraction of data sets.
On Air Hand Gesture Recognition System to Recognize the Digits using an accelerometer and KNN -DTW
• We present an efficient On Air Hand Gesture Recognition for Digits method based on an accelerometer Sensor using K-Nearest Neighbor (KNN) and Dynamic Time Warping (DTW). This application requires a training sample per digit gesture.
• We show that there are considerable variations in gestures collected over a time and in gestures collected from multiple users.
• We report an extensive evaluation of digit data with over 3200 gesture samples collected from 6 users of a predefined vocabulary of 10 gesture patterns.