🦻 Hearing Self-Test
Adaptive staircase audiometry: wear earphones in a quiet environment. System plays pure tones at 8 frequencies. Press and hold when you hear, release when you don't. Two rounds per ear.
Calibration Tone
1000Hz reference tone before test starts. Use phone volume keys to set a comfortable speaking level. Then the test begins automatically.
inheritTest
When creating a preset with inheritTest enabled, the EQ values are stacked on top of the self-test compensation curve instead of starting from zero.
🎛 Equalizer (EQ)
5-band graphic equalizer with independent left/right ear adjustment. Each band covers a specific frequency range.
≤600Hz (b0)
Low-frequency rumble zone. Reduce to minimize AC hum, road noise. Range -24~+6dB.
~1.2kHz (b1)
Mid-low transition. Affects body and warmth of sound. Range -24~+6dB.
~2.4kHz (b2)
Speech core zone. Most vocal energy concentrates here. Directly affects clarity. Range ±12dB.
~4kHz (b3)
Sibilance zone. Boost adds detail and air, but too much causes harsh "hissing". Range ±12dB.
≥4kHz (b4)
High treble zone. Affects sparkle and spaciousness. Elderly typically have reduced sensitivity here. Range ±10dB.
💾 EQ Presets
Save current EQ slider positions as named presets. Switch anytime. Supports clipboard import/export for sharing.
💾 Save: Save current slider positions as a new preset or overwrite existing one.
📤 Export: Copy current preset as JSON text to clipboard for sharing or archiving.
📋 Import: Read JSON preset from clipboard and add to list. Auto-handles name conflicts.
🗑 Delete: Remove current preset. Cannot delete the last preset (fallback auto-created).
Export format: {"eq_preset":{"name":"outdoor","b0":-24,"b1":-12,"b2":8,"b3":0,"b4":0}}
📊 Gain & Compression (WDRC)
Wide Dynamic Range Compression. Amplifies quiet sounds while preventing loud sounds from becoming harsh.
Gain Style
5 gain levels: Quiet(-3dB), Elegant(0dB), Loud(+6dB), Powerful(+9dB), Bold(+12dB). Higher values increase overall loudness but may introduce noise floor.
Smooth Mode
Controls how fast gain changes. Fast (better transients), Slow (stable background), Adaptive (dynamic switching, recommended).
Smooth Duration
Time for gain to transition from current to target. Shorter responds faster but may cause pumping; longer is stable but sluggish.
Expander (Noise Gate)
Non-binary noise gate. Smoothly attenuates noise floor below threshold without hard muting. Scene-dependent parameters.
🔇 Noise Reduction & Feedback Cancellation
Deep Learning NR
RNNoise-based neural network for real-time speech/noise separation. Effective for steady noise (fans, AC, road), less so for transient sounds. ●=ON ○=OFF.
Adaptive Feedback Canceller
FIR-NLMS adaptive filter detecting and suppressing howling in real-time. Short feedback path (64 taps @ 16kHz = 4ms). Reduced step size to prevent divergence from WDRC nonlinearity. ●=ON ○=OFF.
⚙️ Pre-Noise-Processing Modes
Pre-processing applied before deep learning NR. Four modes affect noise reduction quality and sound character.
T001 Road Noise
Low-shelf filtering + crest limiting. Suitable for traffic/wind noise. Low-shelf cuts rumble, crest limiter tames transients.
Adjustable: low-shelf freq (80~240Hz), gain (0~-6dB), crest threshold (3~53dB), max attenuation (0~24dB)
T002 Crest Limiting
Crest limiting only, no low-shelf. Best for impulsive sounds (door slams, dish clatter). Preserves low-end integrity.
T003 Combined
Low-shelf + crest limiting simultaneously. Most comprehensive pre-processing for complex noise environments. Note: most low-end loss.
OFF
Raw audio straight to deep learning NR. Best sound quality but weakest NR. Recommended only in quiet environments.
🌍 Scene Detection
Audio features + phone sensor fusion to detect indoor/outdoor/crowd scenes. Automatically switches WDRC parameters.
Indoor
Default mode. Gentle compression, ample gain, preserves vocal richness. HPF disabled.
Outdoor
Low gain(-3dB) + strong compression + 230Hz HPF. Suppresses wind/road noise. Higher expander threshold (-45dBFS).
Crowd
Very low compression knee + high compression ratio + fast attack. Babble noise compressed immediately. HPF enabled. Forces kAdaptive smooth mode.
