Method of Proficient Typing Using a Limited Number of Classes (Case No. 2024-063)

Summary:

UCLA researchers in the Department of Electrical and Computer Engineering have developed a novel software algorithm to rapidly predict text using small keyboards for various applications, including mobile computing, gaming, and human-computer interactions. 

Background:

Advancements in mobile computing have drastically changed everyday life and created more convenient, connected lifestyles. Mobile computing utilizes portable devices such as smartphones, tablets, and wearables, which are highly dependent on efficient communication schemes with its users. In the realm of typing, efforts have been made to use smaller keyboards with less than 26 keys, for example, T9 typing. Implementing these schemes into everyday mobile computing devices, however, has encountered obstacles due to impracticalities. Predictive text and AI innovations have improved these efforts, but these methods still can be challenging to use and require users to manually select desired words, effectively defeating their purported benefits and rendering them obsolete. Existing technologies are hampered by the need to map 30 or more different brain signals from the user to the letters of the English alphabet and some punctuation marks. The utility of current devices is also limited by time and memory constraints. There is an unmet need to develop algorithms that can accurately predict the desired word from these user instructions, so that these efficient typing methods may be implemented in applications such as mobile computing and gaming settings.

Innovation:

UCLA researchers in the Department of Electrical and Computer Engineering have developed a software platform that enables rapid typing in English using only 6 or 9 key keyboards. This system relies on recurrent neural networks, which is a machine learning algorithm that uses inferences from previous datapoints and context to estimate subsequent words. After a character is inserted, this method assigns probability to each upcoming word and presents it to the user. This type of neural network is highly effective, as they have achieved near perfect accuracy in typing up to 120 words per minute using 9 key keyboards. The researchers achieve this using both a QWERTY keyboard configuration as well as an alphabetical one. The neural network and keyboard layout can be potentially modified for different languages, further amplifying the versatility of this technology. 

This new software platform can accelerate human communication for both commercial users as well as BCI patients. The implications of this technology extend far beyond conventional communication. In the realm of mobile computing, it offers a transformative solution for enhancing productivity and streamlining text input on smartphones and tablets. Likewise, in gaming applications, where rapid and precise input is crucial, this predictive technology opens new possibilities for immersive and responsive gameplay experiences. Ultimately, this innovative software platform not only accelerates human communication for commercial users but also holds immense potential for improving the quality of life for individuals with disabilities, including those utilizing Brain-Computer Interfaces (BCIs). Its versatility and performance make it a game-changer in the fields of mobile computing and gaming, promising enhanced efficiency, accessibility, and user engagement across various platforms and applications.

Potential Applications: 

•    Human/Brain computer interfaces (BCI/HCI)
•    Gaming
•    Mobile Computing
•    Robotics
•    Consumer electronics
•    Database management/data entry

Advantages: 

•    High accuracy and speed
•    Adaptable to different BCI hardware implementations
•    Can be trained for different languages
•    Compatible with multiple keyboard layouts
•    Utilizes much less keys to encode the alphabet and punctuation (5 to 9 keys)


Development-To-Date:

The researchers have fully designed and successfully demonstrated the invention in September 2023.

Reference:

UCLA Case No. 2024-063

Lead Inventor:

Jonathan Kao
 

Patent Information:
For More Information:
Joel Kehle
Business Development Officer
joel.kehle@tdg.ucla.edu
Inventors:
Jonathan Kao
Shreyas Kaasyap
John Zhou
Johannes Lee
Nima Hadidi