In both user-dependent and user-independent principles, our method outperformed all the state-of-the-art methods by a clear margin for all datasets. In this paper, we extensively investigated different interpolation techniques on seven publicly available air-writing datasets and developed a method to recognize air-written characters using a 2D-CNN model. Interpolation is a statistical technique that can be employed for time-series signals to ensure minimum data loss. To make the signals consistent in length, researchers attempted various strategies including padding and truncating, but these procedures result in significant data loss. Inconsistent signal duration is obvious due to the nature of the writing and data-recording process. However, the variability of signal duration is a key problem in developing an air-writing recognition model. Therefore, air-writing recognition systems are becoming more flexible day by day. Traditional air-writing requires an extra device containing sensor(s), while the wide adoption of smart-bands eliminates the requirement of the extra device. Rather, it is sensitive to the subject and language of interest. Air-writing, unlike general gestures, does not require the memorization of predefined special gesture patterns. Air-writing recognition is a special case of gesture recognition in which gestures correspond to characters and digits written in the air. The act of writing letters or words in free space with body movements is known as air-writing. Aided by the tolerance of character ambiguity and accurate character recognition, SHOW achieves over 70% lower mis-recognition-rate, 43% lower no-response-rate in both daily and general purposed text-entry scenarios, and 33.3% higher word suggestion coverage than the tap-on-screen method using a virtual QWERTY keyboard. Furthermore, having more screen space after removing the virtual keyboard, SHOW can display 4x candidate words for autocompletion. Our experiments show that SHOW can effectively generate 60 traces from one real handwriting trace and achieve high accuracy at 99.9% when recognizing the 62 different characters written by 10 volunteers. it does not require whole-arm posture, hence is better suited to space-limited places (e.g. it employs a novel rotation injection technique to substantially reduce the effort of data collection 2. SHOW differs from previous work of gesture recognition in that: 1. SHOW captures the gyroscope and accelerometer traces and deduces the user's handwriting thereafter. In this paper, we propose SHOW, which enables the user to input as they handwrite on horizontal surfaces, and the only requirement is to use the elbow as the support point. Voice input is strongly constrained by the surroundings and suffers from privacy leak. Tap-on-screen is error prone due to the small screen 2. With the two de facto solutions: tap-on-screen and voice input, text-entry on the watch remains a tedious task because 1. Smart watch is becoming a new gateway through which people stay connected and track everyday activities, and text-entry on it is becoming a frequent need.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |