Epiq Solutions
The existing Finite Impulse Response (FIR) filter design routines in the DSP.jl package are window based - a suboptimal technique that starts with an optimal L2 approximation to a desired response, and then applies a window to smooth out the ringing around the transition band. “remez” - the common name for the Parks-McClellan algorithm - is a better technique that gives much better control over the filter’s deviations from the desired response in the pass- and stop-bands. It designs FIR filters with optimal Chebyshev approximations to the desired response. In this talk I’ll provide an overview of the algorithm and efforts to include a pure Julia implementation in DSP.jl, and compare it to scipy’s C implementation.
I’ve been working in signal processing and wireless communications, doing software and system development for over 20 years. I worked 5 years at MathWorks (makers of MATLAB) between my MS and PhD, and 15+ years combined at Motorola and Epiq Solutions. I am coinventor on 15 issued US patents and have authored or coauthored over 20 conference and journal papers. I am on linked in and github. I gave a talk at my son’s High School AP CS class (click on a slide to zoom in, arrow keys to navigate) about computers and engineering that also includes some of my personal philosophy about learning and life.
