Zengli Yang

Generative AI, ASR, TTS, AI Assistant, VLM, CV, Video Understanding, Multimodality

AI, computer vision and deep learning, including generative AI, llm, camera calibration, vehicle surround view, vehicle security, etc.

Algorithm research and development on computational photography and computer vision. o Image and video semantic segmentation and refinement o Stereo depth estimation and refinement o Deep learning for image denoising and demosaicing o Camera ISP development and support, including denoising, demosaicing, sharpening, super-resolution, etc. o Automatic log analysis with natural language processing (NLP)

4G LTE Modem firmware development, simulation, emulation, verification, silicon bring-up, and on-target support

Adaptive Turbo Equalization in MIMO Underwater Acoustic Communications o Proposed iterative channel estimation based adaptive turbo equalization algorithms for large MIMO time-varying channel o Processed ocean experimental data and verified proposed algorithms with publications

DSP Implementation of Underwater Wireless Communication System o Designed/Simulated the wireless communication algorithms in the physical layer using the techniques of direct-sequence spread spectrum, adaptive equalization, TDOA localization, etc. o Implemented low power underwater acoustic transceiver and network for accurate localization and robust communication based on TI TMS320C6713 DSP o Integrated the acoustic transducers, power amplifier, low-noise amplifier, GPS, ADC, DAC, etc. o Performed comprehensive test and verification in lab and field experiments

FPGA Implementation of MMSE Detector for Turbo MIMO-OFDM Transceiver o Designed and developed resource efficient MMSE detector for MIMO-OFDM system using Verilog HDL, and decreased the resource usage on Altera Stratix III and Stratix IV FPGA by 50% using systolic array o Performed system test, verification, debugging, integration, and improved the performance in one month

Compressed Sensing for 3-D Imaging Systems o Developed efficient 3-D radar imaging algorithms with undersampled measurements by utilizing compressed sensing with l1-norm convex optimization and image processing techniques o Compared the performance of the sparse 3-D near-field SAR and MRI imaging methods using compressed sensing techniques qualitatively and quantitatively

Robust Adaptive Algorithms for Echo Cancellation in Impulsive Interference Environment o Proposed efficient algorithms robust against impulsive interference and channel with different sparseness for audio signal enhanced sparse system identification and network echo cancellation

Log Visualization Tool for Voice Processing Module o Developed visualization tool using Python and MS Visio library to help debug and monitor the audio console message o Reduced debug time by 98% and achieved high evaluation from the manager/team