(Eldorado Research Institute)
📅Jul 01, 2019
To improve speech recognition for people with distinct accents, we train an AI algorithm to recognize the speech of those said people in particular. At first, we provide a generic speech recognition device that operates isolated. If a user notice that the device fails to recognize some words, he can point out the mistake to the device and input the correct interpretation of his words using a keyboard. The device will learn from this correction and improve its speech recognition algorithm's efficiency for that specific user. This specialized algorithm for that user can be shared with others users that have a similar accent or other devices operated by that user.
(California Polytechnic University)
📅Jun 30, 2019
A motion controller is a control system in charge of the moving parts of a machine. A motion controller can range from a simple fan controller to multiple axis CNC machines, 3D printers robots, and more. Industrial motion controllers are usually application specific and cannot be easily re-configured to suit multiple purposes, as well as being hard to obtain for the maker community and closed-source. Open-source motion controllers have their own slew of disadvantages, mainly stemming from being micro-processor based which results in the controller being only able to do one task at a time which limits pulse generation and the number of axises, as well as a distinct lack of safety measures. These controllers also traditionally use a trapezoidal motion profile which tends to create high oscillation in the movement of the controlled device. The main goal of this project is to create an open-source motion controller that can solve most of the problems of both industrial and open-source motion controllers by implementing different features and using a kinematic based, optimized motion profile with the S-curve algorithm. An S-curve’s defining point features two acceleration periods, followed by a period of no acceleration at its max velocity, and finally two decreasing deceleration period. Compared to a trapezoidal motion algorithm that uses linearly increasing acceleration and deceleration, the S-curve motion controller has much less jerking (oscillating) movement which results in smoother motion overall.
📅Jun 30, 2019
📅Jul 01, 2019
EEG signals are used to build mode of game and will be used in future VR/AR. We can have multiple applications . We are using FPGA Nano kit to implement .
📅Jul 02, 2019
Industrial drone auto piloting, waypoint mapping and navigation using multiple inferences running concurrently on the edge device, using OpenVINO on main board(s) and couple of Movidius sticks with pretrained networks. End result will be actual drone flying through the arbitrary obstacle course similar to real-world industrial environment. Big part of this effort is aimed towards safety (human interaction, dmz zones etc.).