berkus
/
wifi-densepose
mirror of https://github.com/ruvnet/wifi-densepose.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
wifi-densepose/v2/crates/wifi-densepose-signal
History
rUv 8c24b8bdfe
refactor(beyond-sota): ADR-154 M3 — clear §7.4 P3 backlog (22 de-magic + 6 boundary tests, backlog 36→0) (#1057) 
* refactor(signal): de-magic motion.rs tuning constants (ADR-154 §7.4 #18)

Lift the bare fusion weights, normalization scales, confidence-indicator
weights, and adaptive-threshold clamp bounds in motion.rs out of the
scoring functions into named, documented EMPIRICAL-DEFAULT consts. Values
are bit-identical to the prior literals — this is cleanup, no behaviour
change.

Adds boundary/characterization tests pinning current behaviour:
- motion_tuning_consts_unchanged_from_literals (consts == old literals)
- doppler_component_saturates_at_full_scale (/100 then clamp(0,1))
- correlation_score_zero_below_n2_boundary (n<2 guard)
- temporal_variance_zero_below_two_history (len<2 guard)
- adaptive_threshold_engages_at_history_boundary (history 9 vs 10)

Co-Authored-By: claude-flow <ruv@ruv.net>

* refactor(signal): gesture.rs euclidean length guard + de-magic (ADR-154 §7.4 #12)

- Add a debug_assert! to euclidean_distance documenting the same-dimension
  caller contract: zip() silently truncates on a length mismatch, so a
  mismatch is now loud in debug builds while the release operating path and
  output are unchanged.
- De-magic the bare 1e-10 confidence epsilon into a documented const
  CONFIDENCE_SECOND_BEST_EPSILON (value unchanged).

Tests pinning current behaviour:
- confidence_epsilon_unchanged_from_literal
- dtw_empty_sequence_is_infinite (n=0/m=0 boundary)
- euclidean_distance_equal_length_is_l2 (same-dim contract)

Co-Authored-By: claude-flow <ruv@ruv.net>

* refactor(signal): de-magic longitudinal.rs drift thresholds (ADR-154 §7.4)

Lift the bare drift-detection literals (7-day baseline, 2-sigma z-score,
3-day sustained, 7-day escalation, EMA alpha, cosine epsilon) into named,
documented EMPIRICAL-DEFAULT consts encoding the module's Key Invariants.
The duplicated `>= 7` in is_ready/is_ready_at now share one const. EMA alpha
kept as the exact 0.05 literal (1.0 - 0.95_f32 is not bit-identical in f32).
Values unchanged.

Tests:
- drift_consts_unchanged_from_literals
- is_ready_at_day_boundary (day 6 vs 7)
- cosine_similarity_zero_vector_is_zero (zero-norm guard)

Co-Authored-By: claude-flow <ruv@ruv.net>

* refactor(signal): de-magic division/zero-norm epsilons + boundary tests (ADR-154 §7.4)

De-magic the bare division-guard epsilons in four modules into named,
documented consts (values unchanged) and pin the previously-untested
zero-norm / zero-variance / degenerate boundaries:

- cross_room.rs: COSINE_SIMILARITY_EPSILON (1e-9) + test_cosine_similarity_zero_vector
- multiband.rs: PEARSON_DENOMINATOR_EPSILON (1e-12) + pearson_correlation_zero_variance
- intention.rs: LEAD_TIME_MIN_ACCEL (1e-10) + lead_time_zero_for_static_stream
- hampel.rs: ZERO_MAD_EPSILON (1e-15) + test_zero_half_window_error
  + test_zero_mad_constant_window; documented hampel_filter # Errors

Each module also gets a *_unchanged_from_literal const-pin test.

Co-Authored-By: claude-flow <ruv@ruv.net>

* refactor(signal): de-magic rf_slam + attractor_drift constants (ADR-154 §7.4)

rf_slam.rs:
- NS_PER_DAY (86_400_000_000_000.0), MIGRATION_MIN_SPAN_DAYS (1e-9), and the
  fixed-map defaults (FIXED_MAP_ASSOC_RADIUS_M/MIN_SIGHTINGS/MIN_COHERENCE)
  lifted out of inline literals (values unchanged).
- migration_zero_span_is_zero_rate pins the single-sighting zero-span guard.

attractor_drift.rs:
- METRIC_BUFFER_CAPACITY (365), STABLE_CENTER_WINDOW (10) de-magicked.
- Documented the implicit recent.len()>=1 divide-safety in the PointAttractor
  branch (guaranteed by the count < min_observations guard).
- analyze_min_observations_boundary pins the off-by-one boundary.

Each module gets a *_consts_unchanged_from_literals pin test.

Co-Authored-By: claude-flow <ruv@ruv.net>

* refactor(signal): de-magic coherence.rs variance floor + default decay (ADR-154 §7.4)

Completes the M1 #9 de-magic for coherence.rs: the four bare 1e-6 variance-floor
literals (update_reference floor + coherence_score/per_subcarrier_zscores epsilon)
collapse to one VARIANCE_FLOOR const, and the inline 0.95 default decay becomes
DEFAULT_EMA_DECAY. Values unchanged.

Tests:
- drift_consts_unchanged_from_literals extended (VARIANCE_FLOOR, DEFAULT_EMA_DECAY)
- coherence_score_finite_with_zero_variance pins the floor's effect

Co-Authored-By: claude-flow <ruv@ruv.net>

* refactor(signal): de-magic calibration.rs thresholds + min-frames default (ADR-154 §7.4 #2)

Lift the bare calibration literals into named EMPIRICAL-DEFAULT consts (values
unchanged, bit-identical; calibration is off the Python proof path):
- DEFAULT_MIN_FRAMES (600) — was repeated across all four tier constructors
- AMP_STD_FLOOR (1e-12) z-score divisor floor
- MOTION_AMP_Z_THRESHOLD (2.0) / MOTION_PHASE_DRIFT_THRESHOLD (π/6) — the two
  motion_flagged sites now share one definition
- SUBTRACT_MIN_NORM (1e-30) baseline-subtraction guard

Test calibration_consts_unchanged_from_literals pins all five and asserts every
tier constructor shares DEFAULT_MIN_FRAMES.

Co-Authored-By: claude-flow <ruv@ruv.net>

* refactor(signal): de-magic fusion_quality + temporal_gesture constants (ADR-154 §7.4)

fusion_quality.rs:
- CONTRADICTION_PENALTY (0.8) and CONTRADICTION_BOUND_HALFWIDTH (0.1) named.
- no_contradiction_is_identity pins the n=0 boundary (penalty 0.8^0 = 1.0,
  zero-width bounds).

temporal_gesture.rs:
- CONFIDENCE_SECOND_BEST_EPSILON (1e-10, mirrors gesture.rs) and
  NORM_QUANTIZATION_SCALE (1000.0) named.

Each module gets a *_consts_unchanged_from_literals pin test. Values unchanged.

Co-Authored-By: claude-flow <ruv@ruv.net>

* docs(adr-154): record Milestone-3 — §7.4 row #21-45 P3 backlog cleared

Replace the lumped #21-45 backlog row with the enumerated M3 resolution: 22
magic constants de-magicked into named EMPIRICAL-DEFAULT consts (each pinned ==
prior literal), 6 boundary/characterization tests, ~4 doc-only, across 11
modules; not-real findings reported + skipped (unreachable attractor_drift
div0, non-existent gesture thresholds, proof-path features.rs). Update residual
P3 rows #2/#12/#17/#18 to RESOLVED, the deferred count (36 -> 0), the scope
field, and the Horizon-ledger one-liner. §7.4 backlog fully cleared across
M0-M3. CHANGELOG [Unreleased] entry added.

Validation: signal lib --no-default-features 476/0/1; --features cir 476/0;
workspace 3,275/0; Python proof PASS, hash f8e76f21...46f7a UNCHANGED.

Co-Authored-By: claude-flow <ruv@ruv.net>

---------

Co-authored-by: ruv <ruvnet@gmail.com>
 		2026-06-13 19:36:05 -04:00
..
benches perf(beyond-sota): ADR-154 M2 — FFT planner hoist (1.84x, bit-identical) + 3 honest perf nulls + boundary tests (#1055) 2026-06-13 17:34:37 -04:00
src refactor(beyond-sota): ADR-154 M3 — clear §7.4 P3 backlog (22 de-magic + 6 boundary tests, backlog 36→0) (#1057) 2026-06-13 19:36:05 -04:00
tests fix(csi-ingest): real HE20 CSI no longer dropped or replaced with simulated data (#1009, #1004) 2026-06-12 16:37:55 -04:00
Cargo.toml perf(beyond-sota): ADR-154 M2 — FFT planner hoist (1.84x, bit-identical) + 3 honest perf nulls + boundary tests (#1055) 2026-06-13 17:34:37 -04:00
README.md chore(repo): rename rust-port/wifi-densepose-rs → v2/ (flatten to one level) (#427) 2026-04-25 21:28:13 -04:00

README.md

wifi-densepose-signal

Crates.io Documentation License

State-of-the-art WiFi CSI signal processing for human pose estimation.

Overview

wifi-densepose-signal implements six peer-reviewed signal processing algorithms that extract human motion features from raw WiFi Channel State Information (CSI). Each algorithm is traced back to its original publication and integrated with the ruvector family of crates for high-performance graph and attention operations.

Algorithms

Algorithm Module Reference
Conjugate Multiplication csi_ratio SpotFi, SIGCOMM 2015
Hampel Filter hampel WiGest, 2015
Fresnel Zone Model fresnel FarSense, MobiCom 2019
CSI Spectrogram spectrogram Common in WiFi sensing literature since 2018
Subcarrier Selection subcarrier_selection WiDance, MobiCom 2017
Body Velocity Profile (BVP) bvp Widar 3.0, MobiSys 2019

Features

  • CSI preprocessing -- Noise removal, windowing, normalization via CsiProcessor.
  • Phase sanitization -- Unwrapping, outlier removal, and smoothing via PhaseSanitizer.
  • Feature extraction -- Amplitude, phase, correlation, Doppler, and PSD features.
  • Motion detection -- Human presence detection with confidence scoring via MotionDetector.
  • ruvector integration -- Graph min-cut (person matching), attention mechanisms (antenna and spatial attention), and sparse solvers (subcarrier interpolation).

Quick Start

use wifi_densepose_signal::{
    CsiProcessor, CsiProcessorConfig,
    PhaseSanitizer, PhaseSanitizerConfig,
    MotionDetector,
};

// Configure and create a CSI processor
let config = CsiProcessorConfig::builder()
    .sampling_rate(1000.0)
    .window_size(256)
    .overlap(0.5)
    .noise_threshold(-30.0)
    .build();

let processor = CsiProcessor::new(config);

Architecture

wifi-densepose-signal/src/
  lib.rs                 -- Re-exports, SignalError, prelude
  bvp.rs                 -- Body Velocity Profile (Widar 3.0)
  csi_processor.rs       -- Core preprocessing pipeline
  csi_ratio.rs           -- Conjugate multiplication (SpotFi)
  features.rs            -- Amplitude/phase/Doppler/PSD feature extraction
  fresnel.rs             -- Fresnel zone diffraction model
  hampel.rs              -- Hampel outlier filter
  motion.rs              -- Motion and human presence detection
  phase_sanitizer.rs     -- Phase unwrapping and sanitization
  spectrogram.rs         -- Time-frequency CSI spectrograms
  subcarrier_selection.rs -- Variance-based subcarrier selection

Related Crates

Crate Role
wifi-densepose-core Foundation types and traits
ruvector-mincut Graph min-cut for person matching
ruvector-attn-mincut Attention-weighted min-cut
ruvector-attention Spatial attention for CSI
ruvector-solver Sparse interpolation solver

License

MIT OR Apache-2.0