Hoi Wong
358 Knollglen, Irvine, CA 92614
Phone: (650) 804-5024
Email:
Education / Engineer and M.S. Electrical EngineeringStanford University
B.S. Electrical Engineering and Mathematics (Hons)
University of Wisconsin – Madison
Strengths / - Solid background in signal processing, applied mathematics and software development
- Proficient in C and advanced uses of MATLAB (data structures, OOP, IPCs, GUI, etc.)
- Broad exposure to multiple technical disciplines (software, math, algorithms, electronics)
- Communication skills sharpened through teaching experience
- Align engineering requirements with business objectives to ensure valuable work is done
- Flexible for startup and mature environments
Startup / - From independent research to rapid prototyping.
- Fill in multiple roles before dedicated work function is established
Mature / - Teamwork skills within and across departments
- Broad knowledge allows tight workflow integration, thus improving outcome and reducing potential conflicts
- Writes clean and easy to maintain code through gathering and analyzing true requirements, developing neat architecture by effective use of programming strategies and language features, guided by minimizing total development time factoring in debugging and maintenance costs
Thesis (Engr.) / Macro Adaptive Algorithms (Advisor: Bernard Widrow)
Invented 1) a family of new adaptive algorithms by leveraging the LMS to outperform itself in terms of convergence performance, stability, and robustness with minimal increase in computation time. 2) a new adaptive algorithm structure allowing LMS learning rate to be cranked up for fast convergence way beyond stability limits. Combined they close the gap between LMS and RLS algorithms.
Derived mathematical properties of the new algorithms and explained over-adaptation effects. Results simulated using GPU parallel computing (Accelereyes Jacket) coded in MATLAB classes (OOP) and CUDA C++ (with thrust library).
Publication / H. Wong and W. A. Sethares, “Estimation of Pseudo-periodic Signals”, IEEE Proceedings of Int’l Conference on Acoustics, Speech and Signal Processing, Montreal, May 2004
Experience / Signal Processing Software Design Engineer, Masimo/Cercacor Laboratories
Advanced Algorithm Engineer, Masimo Corporation (Transfer) [Fall 2010-Present]
Signal Processing and Mathematical Algorithms
Core algorithm development for non-invasive total hemoglobin (e.g. to detect anemia and internal bleeding) measurement device: from clinical raw data, signal processing, all the way to final calibration curve (similar to SpO2, non-invasive blood oxygen saturation).
Rewritten core algorithm code chains (at both Masimo Corporation and Cercacor Laboratories) using new data structures (dataset/table/categorical objects) and concepts from relational databases, resulting in full understanding of the mathematical algorithms used and significantly simplified routine experimental work.
Improved non-invasive total hemoglobin (e.g. anemia) measurement accuracy through innovative use of repeated measurements through mathematical optimization.
Discovered a signal processing method to measure venous blood non-invasively while developing mathematical models intended to make non-invasive arterial blood measurements robust against motion artifacts and improve non-invasive total hemoglobin (arterial blood) measurement accuracy.
Proposed and implemented a visualization method similar to Clarke error grid analysis to reinterpret non-invasive total hemoglobin estimation accuracy in terms of clinical significance for non-invasive total hemoglobin.
Algorithm design and development for multiple new non-invasive hemodynamic parameters through optical and acoustic sensors. Developed new mathematical signal models behind fluid responsiveness indicators leading to improvements in artifact rejection and specificity.
Research and development
Exploratory data analysis for non-invasive blood glucose estimation. Implemented pharmacokinetic (PK/PD) state-space models and used Kalman filtering to predict blood glucose levels between finger pricks and non-invasive measurements.
Analyzed trade-offs between hardware requirements and measurement accuracy for the economy product line of non-invasive total hemoglobin sensors. Developed parametric statistical models to increase manufacturing yield (thus reducing costs) while maintaining the same non-invasive hemoglobin measurement accuracy.
Investigated new physiological parameters that can be measured by non-invasive sensor and hardware technologies in multiple product lines within the company, and their applications in perioperative (surgical) monitoring. Responsible for developing prototype analog electronic interface circuits, writing PCB rework specifications, setting up data acquisition systems, experiment design and clinical data collection, all the way to signal processing algorithms and data analysis.
Software (Embedded DSP)
- Took initiative to bring C++ into the new embedded DSP platform for upcoming projects.
Achieved by creating C++ vector and matrix classes (operator overloading and automatic memory management through RAII) for numerical calculations routinely used in signal processing and algorithm development, encapsulating ADI SHARC DSP runtime libraries which I compiled into a proprietary ROM image. The classes greatly simplified the unit tests I wrote to verify the numerical accuracy compared to MATLAB, automated through DCOM over VDSP. In the process, I also mathematically identified and reported a bug in runtime library (VDSP-15997): the DSP runtime library’s coherence implementation included noisy samples in the convolution.
- Brought modern methodologies and language constructs in the embedded software development process
Developed and implemented core algorithms in C on TI OMAP DSP under tight footprint. Extensively used C99/GCC extension features and developed memory leak detection tools to simplify coding and debugging tasks, and introduced the modern language features to the team during code reviews. To help the group meet the code and heap space budget on the embedded DSP, I refactored the algorithm code and managed their memory allocation through TI macros and compiler/linker optimization options to claim a significant amount of headroom
Software (Abstract-level)
Developed compression algorithms for proprietary data formats. Created MATLAB Compiler Runtime-based DLL (C++) solution to interface with compiled LabVIEW program in manufacturing systems. Wrote MEX functions (in C). Written a low level proprietary file reader from scratch based solely on file format specifications document. Written compact libraries to manage inter-process communication (IPCs) to do parallel MATLAB processing without Parallel Computation Toolbox.
Cross-functional
Found hardware root causes of abnormal data patterns in LED characterization data by tailoring probing experiments and analyzing results. Discovered modifications to textbook optical power model which improved the data fitting accuracy.
Business and Market Development
Background due diligence research for a variety of potential business acquisitions. Review and broken down multiple medical device and physiological measurement mechanisms from patents, internal documents, product testing, clinical experiment setup and data analysis. Organized the result based on market segment and measurement modalities for a non-technical audience.
Technology transfer
Developed and delivered presentations to educate multiple teams about the similarities and difference between non-invasive hemoglobin technologies between both companies. Topics ranging from optimization, signal processing, to analog electronics and sensor design. Served as a bridge between the staff of the two companies in the transfer process.
Teamwork, managing vendors
Interfaced with clinical data collection team (vendor). By understanding underlying objectives, lifting obsolete constraints, re-specifying project requirements and re-designing clinical protocol, logistics coordinated with modern software tools, I was able to work together with the team to reduce the human error rate from 12% to <0.003% for extremely costly clinical data.
Given technical presentations for technology transfer, advanced uses of MATLAB, and how analog electronics design (sensors and instrumentation) relate to signal processing algorithm development work.
Software Consultant, Dept of Arrhythmia Services, Stanford School of Medicine [2006-10]
Designed and implemented a system to collect and extract electrocardiogram data from Implantable Cardiac Defibrillators (ICDs) from multi-center studies, store, organize and visualize the data. Developed a user-interactive heart-rate estimation algorithm and annotation tools to aid electrocardiogram scoring.
Worked with multiple vendors (e.g. Guidant, Medtronic) to extract ICD data from proprietary formats. Written XLST scripts to convert document-oriented XML to data-oriented XML. Without access to documentation, reconstructed the specification by reading through a half-written Perl code base. Developed C# and MATLAB GUIs and anonymization algorithms for traceable yet HIPPA-compliant remote data transmission. Written GUI to provide similar features as an ICD telemetry reader.
Research Assistant, Dept of Anesthesia, Stanford School of Medicine [Summer 2007]
Analyzed, acquired and set up data acquisition solution for the laboratory. Quickly learned Igor and developed a graphical user interface for acquiring data.
Bluetooth Software Development Intern, Broadcom Corporation [Summer 2006]
Developed signal quality analysis model to evaluate echo cancellation algorithm in Bluetooth audio. Created objective performance metrics using mathematical techniques and implemented a fully automated measurement setup (signal generation, capture and analysis) using MATLAB-UART interface to the development board.
Bridged the gap between embedded software development team and software testing & verification team (SVT). Gained understanding of the Bluetooth stack and protocol. Developed an automated unit test scripting framework and example scripts for internal WIDCOMM Bluetooth testing program. Debugged and modified script engine written in mixed C and C++. Prepared and delivered presentation to the SVT, the client of the testing package. Debugged firmware on ARM 7-TDMI development system.
Imaging Intern, PortalPlayer [Summer 2005]
Developed digital still camera image processing algorithms (AWB, AGC, AEC) controlling the image sensors on ARM-based embedded systems coded in C/C++. Analyzed algorithms and sensor behavior and set up a studio for verification and testing. Prepared and delivered technical presentations and documented the algorithm development process
Research Assistant, Dept. of Otolaryngology, Stanford School of Medicine [2004-2005]
Developed a signal processing system (simulated baby ear) to automate testing of hearing diagnosis equipment in hospitals that measures Distortion Product Oto-Acoustic Emissions (DPOAE) and Auditory Brainstem Response (ABR)
Designed and prototyped signal conditioning circuit and acquired and processed real-time data using MATLAB DAQ Toolbox. Devised and implemented algorithms on embedded DSP (TMS320) using C.
Teaching
Assistantships
(Stanford) / Spectral Audio Signal Processing / Applications of FFT [EE/Music - Spring 2006, 2008]
Medical Device, Diagnostics, and Pharmaceuticals [BioE - Spring 2006]
Biomedical Electronics Laboratory [EE – Spring 2010]
(Covers Galvanic Isolation, LNA, ECG, ICG, PPG)
Analog Electronics Laboratory [Physics – Fall 2006]
Awards / Vincent C. Rideout Scholarship (2003)
Clarence F. and Helen A. Martin Scholarship (2002)
Pastime / Collect and experiment with (and occasionally repair) old test and measurement equipment
Programmed GPIB commands to control test instruments.
Hardware hacking
Built and programmed low cost audio with Atmel (AVR) PWM output and experiment with NXT speaker technology for toy development.
Built a simple ECG circuits to demonstrate proof-of-concept intra-limb muscle tremor noise cancellation.