<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Edgeai on Javad Rajabzadeh | Senior Software Engineer</title><link>http://javad.dev/tags/edgeai/</link><description>Recent content in Edgeai on Javad Rajabzadeh | Senior Software Engineer</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sun, 19 Jul 2026 22:00:00 +0330</lastBuildDate><atom:link href="http://javad.dev/tags/edgeai/index.xml" rel="self" type="application/rss+xml"/><item><title>Real-Time Multichannel Speech Enhancement: Inverse-Free Rank-N Matrix Updating on Edge Hardware</title><link>http://javad.dev/posts/rankn-multichannel-wiener-filter/</link><pubDate>Sun, 19 Jul 2026 22:00:00 +0330</pubDate><guid>http://javad.dev/posts/rankn-multichannel-wiener-filter/</guid><description>&lt;p&gt;In real-world acoustic engineering—especially on &lt;strong&gt;drones, autonomous robotic platforms, and ultra-low-power edge processors&lt;/strong&gt;—enhancing speech signals in real time presents a severe computational bottleneck. Microphones mounted on moving drones or robotic units encounter heavy spatial interference, rotor harmonics, and turbulent wind noise.&lt;/p&gt;
&lt;p&gt;While classical Multi-channel Wiener Filters (MWF) offer optimal spatial noise reduction, computing full spatial matrix inversions $\mathbf{R}_{nn}^{-1}(f)$ across hundreds of STFT frequency bins imposes an $\mathcal{O}(M^3)$ computational load that degrades battery life and causes latency spikes on embedded microcontrollers (such as STM32, ESP32, or ARM Cortex-M flight controllers).&lt;/p&gt;</description></item></channel></rss>