Researchers Ben Pearman and Anand Kumar have put a LimeSDR Mini 2.0 to use in a very unexpected way: seeing through walls, with the help of an artificial intelligence (AI) model.

“We humans inhabit a world of sight, sound, smell, taste, and touch. We cannot directly perceive the electromagnetic fields powering many of our modern technologies without specialised tools,” Ben explains of the project. “Technologies such as cameras and LiDAR enhance our visual perception but are limited to clear lines of sight. Medical imaging technologies have successfully overcome some line-of-sight constraints but are often limited by factors like high power consumption, large size, and substantial costs.

“Motivated by these limitations, we explored whether the relatively low frequency radio waves from an affordable, consumer-grade SDR (the LimeSDR Mini 2.0) could detect hand gestures and poses through occlusions like walls by leveraging the inherent properties of radio waves in the 2-4GHz range, which experience lower attenuation through common building materials compared to higher-frequency waves.”

Using the LimeSDR Mini 2.0 and feeding its data into a pair of convolutional neural networks (CNNs), the pair were able to classify three hand gestures – a closed fist, an open palm, and a tap – not only with a direct line of sight but through 7.5mm of foam board and even a 45mm-thick wooden block. “This demonstrates the potential of SDRs,” Ben says, “as cost-effective, energy-efficient, and accessible alternatives to traditional high-frequency, multi-antenna radar systems for gesture and pose recognition applications.”

More information is available in Ben’s Field Report on the LimeSDR Mini 2.0 Crowd Supply campaign page; a supporting video is available on YouTube.

Andrew Back has shared an update on the LibreCellular project, which aims to provide a fully-tested and validated hardware platform and software stack for 4G Long Term Evolution (LTE) cellular networking – and which now includes improved hardware and a lower-cost radio front-end powered by a Raspberry Pi RP2040 microcontroller.

“Last year we introduced the first revision of the LibreCellular 2×2 MIMO RF Front-End (RFE),” Andrew explains. “When we came to characterise the RFE we came to the conclusion that there was room for improving the RF performance, given that the gain we were measuring when switching amplifiers out of bypass was somewhat below the stated figures for the devices used.”

In experimenting with the redesign, the team cam up with the LibreCellular RF Front-End Mini – a cost-reduced and easier-to-debug alternative to the RFE, which features a Raspberry Pi Pico and its RP2040 microcontroller on-board. Proving the intended changes in this smaller design, the full LibreCellular RF Front-End was also revised to deliver improved performance – though the discovery of excessive insertion loss means another design pass may be required.

Other news from the project includes an update to the Band 3 Medium Power Amplifier hardware, a look at the fully-assembled CSRAN1, a major firmware rewrite and improved graphical user interface, the publication of an Ansible collection, and a “technology primer” which, Andrew explains, “in no way attempts to be a comprehensive guide, but hopefully provides sufficient information for those new to the technologies involved to more quickly get up and running with their own network.”

More information is available in Andrew’s project update.

Del Hatch has designed a fully stand-alone AM/FM software-defined radio, capable of operating without being connected to a PC – and featuring a slick physical interface in robust housing.

“This is a stand-alone AM/FM tuner based on a software define[d] radio (SDR),” Del explains of the project. “It has a graphical display showing the radio signal at various points in the radio processing chain, such as at the RF receiver output as it enters the intermediate filter (IF), the output of the IF filter, the full baseband audio signal just after demodulation, and an audio scope for the L/R audio signal.

“It provides the user the ability to modify the width and filter slope of the IF filter, which is important for selecting the best RF signal while rejecting as much noise as possible. The user can also adjust the ‘high blend’ frequency, which can be used to optimise the audio in marginal reception conditions.”

The trick, of course, is that the SDR is stand-alone only because it brings its own PC along for the ride: a Raspberry Pi 5 single-board computer, hidden behind an attractive front-panel with a custom user interface board and five-inch AMOLED touchscreen display. “The gnuradio-companion software takes the RF data from the SDR,” Del explains, “and processes it in various ways: filtering, down-sampling, AM and FM demodulation.”

The project is documented on Hackaday.io, with a supporting video available on YouTube; source code has been published on GitHub under the permissive MIT licence, along with hardware schematics and files for fabricating the custom PCB.

Maker Mirko Pavleski has put together a shortwave single side-band (SSB) software-defined radio, which digs into radio’s history to feature a low-voltage vacuum tube in its design.

“A vacuum tube radio, also known as a valve radio, is an early type of radio receiver that uses vacuum tubes to amplify and detect radio signals,” Mirko writes by way of introduction. “These radios were widely used from the early 20th century until the 1960s. Nowadays with just one vacuum tube and signal generator you can make a super simple, yet sensitive and selective shortwave/SSB radio receiver, thanks to the use of PC and SDR processing software.

“In fact, this project represents a kind of software-defined radio where components that have been traditionally implemented in hardware (e.g., mixers, filters, amplifiers, modulators/demodulators, detectors, etc.) are instead implemented by means of software on a personal computer or embedded system. This allows for more flexibility and the ability to handle a wide range of frequencies and protocols simply by updating the software.”

The radio is built around a low-cost low-voltage 6J1 vacuum tube, and is a remix of an earlier DRM direct mixer design from Burkhard Kainka. “During the testing I came up with the idea to reduce the supply voltage,” Mirko adds, “and I came up with an amazing result: the receiver works perfectly well even at a voltage lower than 3.7V, so I decided to power it with a single lithium battery.”

The project is documented in full, including schematics, on Hackaday.io, with a supporting video available on YouTube.

RTL-SDR.COM has brought our attention to a project offering real-time automated speech-to-text transcription for radio communication logging: RadioTranscriptor by pseudonymous developer “theckid.”

“RadioTranscriptor [is] a voice transcriptor for SDR,” theckid explains of the project. “Using GPU as AI [Artificial Intelligence] transcriptor ([with] CUDA if available), defaults to CPU usage instead if CUDA is not available. Works best with 48kHz input, automatically resampled to 16kHz for Whisper compatibility.”

The tool is designed around OpenAI’s Whisper model, which runs on-device when given sufficiently powerful hardware – and the performance of which can be accelerated by offloading its work to an Nvidia graphics processor via CUDA. Sampled speech is automatically transcribed to text and saved to a log file, though with a few caveats common to unreliable large language model (LLM) technology: “[the] transcriber does not always capture voice and comes up blank in the logs,” theckid admits, “[while] repeated words are recorded.”

The project’s source code is available on GitHub under the MIT license.

Justin Lutz has published a guide on receiving image data from NASA satellites and then feeding them to an object detection model to automatically identify and locate rain clouds.

“I read recently that NASA weather satellites transmit their data unencrypted, so you can connect to and download imagery data directly from their satellites,” Justin explains. “I also learned that there is one satellite (GOES-19) that is parked in geostationary orbit (22,000 miles away!) at an elevation and azimuth that is viewable from my backyard. In addition, GOES-19 is the latest NASA weather satellite, activated in April 2025, so it has the latest technology on it.

“I bought a kit on Amazon that includes the L-band antenna dish (1.7GHz centre frequency), a low-noise amplifier (LNA), a software defined radio (SDR), and RF SMA cable. All I had to do was provide my Raspberry Pi 4 and an inexpensive tripod to mount the dish to. Once I started collecting data and weatherised the sensitive electronics (the Raspberry Pi, the LNA, and the SDR), I could start collecting lots of data both day and night.”

The gathered data was then fed into Edge Impulse Studio, a platform for training and deploying machine learning models – in this case, training it on how to recognise the appearance of rain clouds in NASA’s image data. “My goal was to create an object detection model to identify rain clouds in my downloaded images,” Lutz explains, “then animate the images into a GIF and display that on a local webserver. I could then be notified if rain clouds were approaching my house.”

The project is documented in full on Hackster.io.

Finally, radio ham and developer Alex “VE3NEA” Shovkoplyas has released an open-source software-defined radio tool which is specifically aimed at satellite tracking and reception: SkyRoof.

“SkyRoof combines satellite tracking and SDR functions in one application, which opens some interesting possibilities,” Alex explains of his software. “For example, all satellite traces on the waterfall are labelled with satellite names, the boundaries of the transponder segments follow the Doppler shift, and all frequency tuning is done visually, with a mouse. The program can work without an SDR, or even without any radio at all, but many useful functions are not available in this mode.”

The satellite-specific tool delivers quick access to information on all known satellites transmitting in amateur bands, provides real-time satellite tracking with pass prediction, provides a “sky view” from the user’s current location, an “Earth view” from the perspective of the satellite, a timeline, and a pass list, offers a waterfall display, acts as an SSB/CW/FM receiver with RIT and Doppler tracking with audio and I/Q output, can control external transceivers via CAT, and even handle antenna rotator control.

More information on SkyRoof is available on the official website, while the source code is available on GitHub under the reciprocal GNU Affero General Public Licence Version 3.