San Diego, California, United States
• Designed and implemented interactive user interface in compatibility with a variety of label printing machines. Developed functions and features to execute printing customization requests on bulk. Integrated label printing feature into Wasatch SoftRIP with imagine/color tuning. • Architecture design of Wasatch SoftRIP as well as key functions & features development / integration, including status reporting, job automation, barcode driven workflow, contour image recognition, contour tracing image auto cutting, etc.
• Developed digital imaging processing software drivers for industrial printer from Epson, Canon, HP, etc. • Developed imaging recognition software and contour cut path drivers for Graphtec, Mutoh, etc. • Collaborated with QA engineers on testing automation, test results analysis and feature optimization.
Image and Speech Recognition and Reconstruction A multilayered feed forward neural network model that learns a hierarchy of sparse feature dictionaries is developed. As the layers ascended up, the spatial and/or temporal receptive field sizes of neurons increase. The feed forward weights are learnt from repeating spatial and temporal coincident patterns to encode features. So when applied to edge-detected images or curve from audios, the combinations of these edges that depict parts of shapes or whole generic shapes from images and features can be used to reconstruct unseen(invisible) speech from audios in top layer.