Experimental validation of a Bluetooth-based speech audio broadcasting model

Summary

Measured audio-frame loss and predicted speech quality of a Bluetooth LE broadcast on a cabled hardware testbed with added attenuation and Bluetooth interference, and extended the team's earlier simulation model to many interfering devices.

Key findings

• Speech was broadcast with LC3 at 16 kHz, 10 ms frames and 32 kbps, settings every broadcast receiver must support, between two nRF5340 boards joined by shielded cable.
• Without repeated transmissions, mean frame loss rose from about 6% to about 22% as added attenuation went from 17.5 to 18.5 dB; with four transmissions per frame, loss stayed at or below 0.4%.
• With one busy Bluetooth LE connection nearby, frame loss without repetition was about 2% to 13%, higher when fewer radio channels were in use; with one repetition it fell to 0% on the hardware.
• In simulation, one repetition kept loss acceptable with up to 5 busy connections or 15 keyboards plus 15 mice; 10 busy connections or 20 keyboards plus 20 mice needed two repetitions.
• The authors use about 2% frame loss as a general upper limit before concealment artefacts become clearly audible, but one male-voice recording had almost no low-quality scores at about 5% loss.
• Section IV (Testbed Evaluation), in the paragraphs introducing external Bluetooth interference, uses four Bluetooth channel maps, from no blacklisting to excluding channels overlapping Wi-Fi channels 1, 6 and 11. The experiment adds one Bluetooth LE connection as interference; it does not transmit Wi-Fi traffic. These maps model channel exclusion, not the load of an active venue Wi-Fi network.
• The title page lists Ghent University/imec and Cochlear Technology Centre affiliations. Its first-page footnote states that this paper results from the imec.icon BLE2AV project, financed by imec and Flanders Innovation & Entrepreneurship.
• Table III reports nonzero mean frame loss even with two retransmissions: 0.55% for 10 busy connections and 0.13% for 20 keyboards plus 20 mice.
• Section IV uses a 2% frame-loss threshold from earlier analysis of 20 speech recordings as a general boundary for almost no PESQ scores below 2.0 (Zone C). In the two-recording testbed study, one recording had almost no Zone C scores at 4.97% mean frame loss; the authors stress that the threshold is only a general guide.
• Section IV evaluates two speech recordings with PESQ rather than listeners. Table I reports mean hardware frame loss of 5.68% and 22.11% at 17.5 and 18.5 dB added attenuation with one transmission, versus 0.00% and 0.40% with four transmissions; these are means, not maxima over individual runs.

Limits

Cabled and shielded testbed, not over the air; one hardware platform; quality judged by an instrumental measure (PESQ), not listeners; multi-device results are simulated; no Wi-Fi traffic was transmitted.

Read the original

doi.org/10.1109/PIMRC59610.2024.10817355

Link checked 13 October 2026. We link to the original and do not host copies.

Cited in

• What signal strength is enough at the farthest seat?Open
• How much do crowds, phones and Wi-Fi reduce reliability?Open

Cite this source

The accessed date is when this library last checked the link; use the date you read it. To point readers to this summary instead, link to https://assistivelisteningknowledgebase.org/evidence/baert-2024-pimrc/

Every finding here cites its source, and who funded each source is labelled. This site does not recommend products. Found an error or a missing study? Suggest a correction.