| name | scbe-audio-intent |
| description | Encode and decode intent values as phase-modulated audio waveforms using FFT-based demodulation. Use when working with the audio intent layer, phase encoding/recovery, or the symphonic cipher carrier signal. |
SCBE Audio Intent
Use this skill for phase-modulated intent encoding and decoding in the SCBE-AETHERMOORE audio layer.
Core Mechanism
Intent values ∈ [0, 1] are encoded as phase offsets on a carrier frequency, then recovered via FFT peak analysis.
Encoding (Phase Modulation)
wave(t) = cos(2π · f_carrier · t + 2π · intent) + noise
- f_carrier = 440 Hz (A4)
- Sample rate = 44100 Hz
- Duration = 0.5 seconds
- Gaussian noise σ = 0.1 added for transmission simulation
Decoding (FFT Demodulation)
- Compute FFT of received waveform.
- Find peak index in positive-frequency half.
- Extract phase angle at peak:
angle(FFT[peak_idx]).
- Normalize:
intent_recovered = (phase mod 2π) / 2π.
Audio Constants
| Name | Value | Purpose |
|---|
| CARRIER_FREQ | 440.0 | Base carrier (Hz) |
| SAMPLE_RATE | 44100 | Samples per second |
| DURATION | 0.5 | Waveform length (sec) |
Workflow
- Accept intent value ∈ [0, 1].
- Generate time vector:
linspace(0, DURATION, SAMPLE_RATE * DURATION).
- Produce carrier with phase offset:
cos(2π·440·t + 2π·intent).
- Add Gaussian noise.
- To recover: FFT → peak detection → phase extraction → normalize.
Accuracy Considerations
- Phase recovery is approximate due to noise and FFT bin resolution.
- For governance decisions, the recovered intent is compared against known target (e.g., 0.75) with tolerance ±0.1.
- Higher sample rates or longer durations improve phase resolution.
Integration Points
- The recovered intent feeds into the coherence check in
governance_9d.
- Coherence is HIGH (0.95) when
|intent - 0.75| < 0.1, LOW (0.4) otherwise.
- The audio layer provides a physical-channel encoding for the v2 (Intent Phase) component.
Guardrails
- Intent must be clamped to [0, 1] before encoding.
- Only use positive-frequency half of FFT for peak detection.
- Phase normalization must use modulo 2π before dividing.
- Noise level (σ=0.1) is fixed for reproducibility; do not increase without testing recovery accuracy.