Acoustic Voice Assessment via Telehealth: What Survives the Compression

Videoconferencing platforms are optimized for intelligible conversation, not for high-fidelity acoustic capture. To deliver clear speech under variable bandwidth conditions, every major platform applies a chain of audio processing:

• Lossy compression. Audio is encoded in formats designed to discard "perceptually irrelevant" information—exactly the kind of fine spectral detail that acoustic measures depend on.
• Noise suppression. Algorithms identify and attenuate non-speech components, which can include the breathy or aperiodic energy that defines dysphonic voice.
• Automatic gain control. Quiet passages are amplified and loud passages compressed, distorting the dynamic range that intensity measures depend on.
• Echo cancellation and de-reverberation. These can introduce subtle nonlinear processing that changes spectral content.

Each of these processes is helpful for conversation and harmful for measurement. The result is a signal that sounds like the patient's voice but is acoustically a different object than what the microphone originally captured.

The most systematic study of this question is Weerathunge and colleagues (2021), who recorded 29 voice samples from patients with dysphonia, transmitted each sample through six HIPAA-compliant videoconferencing platforms, and compared the resulting acoustic measures against the original recordings.

The platforms tested were Zoom (with and without audio enhancements), Cisco WebEx, Microsoft Teams, Doxy.me, and VSee Messenger. The acoustic measures included mean F0, F0 variability, F0 range, SPL range, HNR, L/H spectral ratio, and CPPS (sustained vowel and connected speech).

The summary of findings:

Two findings from this work are particularly important for clinical practice. First, Microsoft Teams was the least disruptive platform, while Zoom with enhancements was the most disruptive—even though Zoom is the most commonly used platform for clinical telepractice. Second, ambient noise from the patient's environment was a significant predictor of measurement differences across nearly every metric. The patient's room mattered more than the patient's internet speed.

The fundamental issue with live videoconferencing is that the audio is processed in real time by the platform before reaching the clinician. The signal that arrives is not the signal the patient produced.

The workaround is to bypass the platform entirely for the recording itself. The patient records the voice sample locally on their own device, using a recording app that produces an uncompressed or losslessly compressed file (WAV, FLAC, or Apple Lossless). They then upload the file separately—via secure portal, secure email, or HIPAA-compliant file transfer. The clinician analyzes the uploaded file rather than the live transmission.

This approach has several advantages:

• No platform compression. The signal reaching the analysis software is the signal the microphone captured.
• Headset microphone is feasible. Patients can use a basic headset placed at the validated 2.5–5 cm distance, which dramatically reduces the impact of ambient room noise.
• Quality control before analysis. The clinician can listen to the file before computing metrics and request a re-recording if there is obvious clipping, environmental noise, or off-task production.
• Reproducible across sessions. Each recording is collected with the same device and protocol, eliminating session-to-session platform variability.

Bridge2AI-Voice consortium validation work has reported that smartphone or tablet recordings made with a headset microphone at 2.5–5 cm distance achieve correlations greater than 0.90 with research-grade equipment for CPP and other key acoustic measures. This is the empirical anchor for the asynchronous workflow.

For a clinician integrating acoustic assessment into a telepractice voice caseload, the following workflow balances technical reliability with patient burden:

Asynchronous recording is the right default, but live capture has legitimate clinical use cases when the limitations are understood:

• F0 monitoring during gender-affirming voice therapy. Mean F0 is preserved across platforms, and within-session pitch tracking during practice is valuable for biofeedback.
• Subjective screening when no objective measure is required. If the goal is to hear the voice and rate perceptually, the platform's audio fidelity is sufficient even when its acoustic fidelity is not.
• Initial triage when no asynchronous option exists. A live impression is better than no impression, provided the clinician documents that the assessment was platform-mediated and acoustic measures were not collected.
• Patient education and feedback. Showing a patient their pitch or loudness in real time during a session does not require absolute accuracy; relative change within a session is informative.

The line to draw is between using the platform's audio for clinical interaction and quantifying the platform's audio as if it were a clean recording. The first is fine; the second is not.

Telepractice acoustic data needs to be defensible if it ends up in an outcome report, a peer-reviewed publication, or a third-party payer audit. Three documentation practices help:

• Describe the recording chain explicitly. "Patient recorded sustained /a/ on iPhone 14 with Apple EarPods at approximately 3 cm from the mouth, in their home office, using Voice Memos app at lossless quality, uploaded via [secure portal name]."
• Use the same chain for baseline and follow-up. Session-to-session comparisons require equipment continuity. If the device changes, document it and treat the new baseline as a new reference point.
• Note any compromises. If a measurement was collected over live videoconferencing despite the limitations, document it explicitly: "F0 collected via Zoom; CPPS not collected this session due to platform limitations."

These practices make the methodology auditable and protect the clinician from later questions about whether observed changes reflect real voice change or measurement artifact.