Factors Affecting the Accessibility of Voice Telephony for People with Hearing Loss: Audio Encoding, Network Impairments, Video and Environmental Noise
Linda Kozma-Spytek, Christian Vogler · 2021 · ACM Transactions on Accessible Computing · doi:10.1145/3479160
Summary
This comprehensive study from Gallaudet University investigates how audio quality parameters affect voice telephony accessibility for people with hearing loss. The research combines a survey of 439 individuals with hearing loss and four controlled experiments involving 114 participants with hearing loss (cochlear implant and hearing aid users) plus 12 hearing controls. The survey revealed a significant gap between telephone importance and satisfaction: 77% of respondents made daily voice calls, yet only 23% were satisfied with their hearing device's performance for telephone listening. While 72% could understand all-to-most of what someone said face-to-face in quiet, this dropped to 50% on the telephone. Respondents identified telephone sound quality and environmental noise as the two biggest barriers, emphasizing that "simply making speech louder is not enough." The four experiments systematically tested narrowband (NB, 300-3,400 Hz) versus wideband (WB, 50 Hz-7 kHz) audio, varying bit rates (5.90-23.85 kbps), packet loss levels (0%, 3%, 20%), and the addition of video and environmental noise. Stimuli were processed through actual AMR-NB and AMR-WB codecs used in mobile networks, with AT&T Labs providing realistic packet loss modeling using the Gilbert-Elliott model. Testing used real phones with acoustic coupling to participants' hearing devices.
Key findings
Wideband audio consistently outperformed narrowband across all experiments. In quiet conditions, WB audio improved speech recognition by 10.4 percentage points (65.2% to 75.7%, p < 0.0001) and significantly reduced perceived mental effort. This advantage persisted across different bit rates, packet loss conditions, and even when video was added—though the WB benefit was greater for audio-only listening than audiovisual. Environmental noise severely degraded performance, reducing speech recognition by approximately 25 percentage points for audio-only listening. Video telephony helped mitigate noise effects: with synchronized video available, speech understanding remained above 80% even in noise. However, the benefit of WB over NB audio showed diminishing returns when high-quality video was present. Packet loss had devastating effects, particularly for people with hearing loss. At 20% bursty packet loss (worst-case for unmanaged Wi-Fi), speech recognition for participants with hearing loss dropped to 42.9% for NB and 51.2% for WB—levels the researchers characterized as rendering telephony unusable. Hearing participants maintained usable performance even at 20% packet loss, highlighting the disproportionate impact on people with hearing loss. Higher bit rates improved speech quality ratings for both NB and WB audio. Critically, the researchers demonstrated that the WB advantage comes from extended audio bandwidth, not just higher bit rates: low-pass filtering WB audio to mimic NB frequencies eliminated the quality advantage despite maintaining the higher bit rate.
Relevance
This research provides concrete, actionable guidance for telecommunications policy and system design. The authors argue that wideband audio support should be required rather than merely recommended in accessibility standards like Section 508 and EN 301 549. They note that VoLTE (Voice over LTE) networks already support AMR-WB at minimal additional bandwidth cost, making universal WB deployment technically feasible. For practitioners, the findings highlight that hearing device features alone cannot overcome fundamental audio quality limitations—the signal itself must be improved at the source. This has implications for organizations deploying VoIP systems, video conferencing platforms, and mobile applications: quality-of-service guarantees that minimize packet loss are accessibility requirements, not just performance optimizations. The research also validates the accessibility value of video telephony, particularly in noisy environments where lipreading can compensate for degraded audio. However, the authors caution that video quality and audio-video synchronization are critical—real-world network conditions can degrade the visual channel in ways not tested here. For the 30 million Americans with hearing loss (12.7% of the population, rising to 63% among adults 70+), these technical parameters directly affect social participation, employment, and independence. The findings establish that accessible telecommunications requires attention to infrastructure-level audio encoding decisions, not just endpoint hearing device compatibility.
Tags: hearing loss · telecommunications · audio codecs · cochlear implants · hearing aids · wideband audio · packet loss · VoIP · mobile telephony
Standards referenced: Section 508 · EN 301 549 · 3GPP