One of the more interesting technical developments involves a new digital voice method known as Radio Autoencoder (RADE). Developed as part of the FreeDV project, RADE uses machine-learning techniques and digital signal processing to improve voice communication over HF radio frequencies. According to research information reported by the ARRL and Amateur Radio Digital Communications (ARDC), the system is designed to occupy approximately half the bandwidth of a traditional single-sideband voice signal while producing more natural-sounding speech under suitable conditions.
Researchers report that RADE can operate at signal-to-noise ratios as low as minus 2 decibels, potentially helping operators understand conversations when signals are weak. The system is already drawing attention from radio manufacturers. Certain FlexRadio transceivers can use a downloadable module supporting the technology, while owners of other compatible radios may be able to use the software through a connected computer. Although it is too early to determine whether RADE will become a widely adopted standard, the technology represents another example of how software and artificial intelligence techniques are influencing amateur radio.
