Datasets:
File size: 16,065 Bytes
85407a7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 | #!/usr/bin/env python3
"""Render and validate a static V1 FOA RIR pilot with SoundSpaces 2.0."""
from __future__ import annotations
import argparse
import hashlib
import itertools
import json
import math
from pathlib import Path
import numpy as np
import quaternion # noqa: F401 - import before Habitat to avoid allocator issues
from scipy.signal import butter, sosfilt
import habitat_sim
import habitat_sim.sim
DEFAULT_RENDERER = {
"sampleRate": 32000,
"frequencyBands": 8,
"directSHOrder": 3,
"indirectSHOrder": 1,
"threadCount": 8,
"updateDt": 0.02,
"irTime": 2.5,
"unitScale": 1.0,
"globalVolume": 0.25,
"indirectRayCount": 15000,
"indirectRayDepth": 200,
"sourceRayCount": 1500,
"sourceRayDepth": 10,
"maxDiffractionOrder": 10,
"direct": True,
"indirect": True,
"diffraction": True,
"transmission": True,
"meshSimplification": False,
"temporalCoherence": False,
"enableMaterials": True,
"dumpWaveFiles": False,
"writeIrToFile": False,
}
CALIBRATION_SOURCES = {
# All six sources are exactly 1 m from the receiver so the linear fit is
# not biased by distance attenuation.
"world_pos_x": ([1.0, 1.35, 0.0], [1.0, 0.0, 0.0]),
"world_neg_x": ([-1.0, 1.35, 0.0], [-1.0, 0.0, 0.0]),
"world_pos_y": ([0.0, 2.35, 0.0], [0.0, 1.0, 0.0]),
"world_neg_y": ([0.0, 0.35, 0.0], [0.0, -1.0, 0.0]),
"world_pos_z": ([0.0, 1.35, 1.0], [0.0, 0.0, 1.0]),
"world_neg_z": ([0.0, 1.35, -1.0], [0.0, 0.0, -1.0]),
}
def canonical_json_hash(value) -> str:
blob = json.dumps(value, ensure_ascii=False, sort_keys=True, separators=(",", ":")).encode("utf-8")
return hashlib.sha256(blob).hexdigest()
def file_hash(path: Path) -> str:
digest = hashlib.sha256()
with path.open("rb") as handle:
for chunk in iter(lambda: handle.read(1024 * 1024), b""):
digest.update(chunk)
return digest.hexdigest()
def set_if_present(obj, key, value) -> None:
if not hasattr(obj, key):
raise AttributeError(f"SoundSpaces acoustics config has no required field {key!r}")
setattr(obj, key, value)
def create_simulator(stage_config: Path, dataset_config: Path, material_json: Path, renderer: dict):
backend = habitat_sim.SimulatorConfiguration()
backend.scene_id = str(stage_config)
backend.scene_dataset_config_file = str(dataset_config)
backend.load_semantic_mesh = True
backend.enable_physics = False
if hasattr(backend, "random_seed"):
backend.random_seed = 20260826
agent = habitat_sim.agent.AgentConfiguration()
sim = habitat_sim.Simulator(habitat_sim.Configuration(backend, [agent]))
if not sim.semantic_scene:
sim.close()
raise RuntimeError("Semantic scene was not loaded; refusing to fall back to default acoustic material")
spec = habitat_sim.AudioSensorSpec()
spec.uuid = "audio_sensor"
spec.channelLayout.channelType = habitat_sim.sensor.RLRAudioPropagationChannelLayoutType.Ambisonics
spec.channelLayout.channelCount = 4
spec.position = [0.0, 0.0, 0.0]
for key, value in renderer.items():
set_if_present(spec.acousticsConfig, key, value)
sim.add_sensor(spec)
audio_sensor = sim.get_agent(0)._sensors["audio_sensor"]
audio_sensor.setAudioMaterialsJSON(str(material_json))
return sim
def set_listener_identity(sim, position) -> None:
agent = sim.get_agent(0)
state = agent.get_state()
state.position = np.asarray(position, dtype=np.float32)
state.rotation = np.quaternion(1.0, 0.0, 0.0, 0.0)
state.sensor_states = {}
agent.set_state(state, True)
def render(sim, source_position) -> np.ndarray:
sensor = sim.get_agent(0)._sensors["audio_sensor"]
sensor.setAudioSourceTransform(np.asarray(source_position, dtype=np.float32))
observation = np.asarray(sim.get_sensor_observations()["audio_sensor"], dtype=np.float32)
if observation.ndim != 2:
raise RuntimeError(f"Unexpected RIR shape: {observation.shape}")
rir = observation if observation.shape[0] == 4 else observation.T
if rir.shape[0] != 4 or not np.isfinite(rir).all() or not np.any(rir):
raise RuntimeError(f"Invalid FOA RIR: shape={rir.shape}")
return np.ascontiguousarray(rir, dtype=np.float32)
def direct_sample_vector(rir: np.ndarray) -> tuple[int, np.ndarray]:
energy = np.sum(np.square(rir.astype(np.float64)), axis=0)
peak = int(np.argmax(energy))
lo, hi = max(0, peak - 1), min(rir.shape[1], peak + 2)
local = rir[:, lo:hi]
index = int(np.argmax(np.sum(np.square(local.astype(np.float64)), axis=0)))
return lo + index, local[:, index].astype(np.float64)
def infer_native_foa(calibration_rirs: dict[str, np.ndarray]) -> dict:
directions, amplitudes, peaks = [], [], {}
for name, (_, direction) in CALIBRATION_SOURCES.items():
peak, vector = direct_sample_vector(calibration_rirs[name])
directions.append(direction)
amplitudes.append(vector)
peaks[name] = peak
design = np.column_stack([np.ones(len(directions)), np.asarray(directions, dtype=np.float64)])
amplitude_matrix = np.asarray(amplitudes, dtype=np.float64)
coefficients = np.linalg.lstsq(design, amplitude_matrix, rcond=None)[0].T
scores = []
for channel, row in enumerate(coefficients):
intercept = abs(row[0])
directional = float(np.linalg.norm(row[1:]))
scores.append(intercept / (directional + 1e-12))
w_channel = int(np.argmax(scores))
direction_channels = [channel for channel in range(4) if channel != w_channel]
best = None
for permutation in itertools.permutations(range(3)):
score = sum(abs(coefficients[channel, 1 + axis]) for channel, axis in zip(direction_channels, permutation))
if best is None or score > best[0]:
best = (score, permutation)
assert best is not None
native_by_world_axis = {}
axis_names = ("world_x", "world_y", "world_z")
for channel, axis in zip(direction_channels, best[1]):
coefficient = coefficients[channel, 1 + axis]
native_by_world_axis[axis_names[axis]] = {
"channel": channel,
"polarity": 1 if coefficient >= 0 else -1,
"fit_coefficient": float(coefficient),
}
return {
"method": "six_equal_distance_axis_sources_linear_fit",
"native_normalization": "N3D",
"w_channel": w_channel,
"w_polarity": 1 if coefficients[w_channel, 0] >= 0 else -1,
"world_axis_mapping": native_by_world_axis,
"fit_coefficients_intercept_xyz": coefficients.tolist(),
"direct_peak_samples": peaks,
}
def native_to_ambix_robot(rir: np.ndarray, mapping: dict, yaw_deg: float) -> np.ndarray:
w = mapping["w_polarity"] * rir[mapping["w_channel"]]
world = {}
for axis in ("world_x", "world_y", "world_z"):
item = mapping["world_axis_mapping"][axis]
world[axis] = item["polarity"] * rir[item["channel"]] / math.sqrt(3.0)
yaw = math.radians(yaw_deg)
cos_yaw, sin_yaw = math.cos(yaw), math.sin(yaw)
right = cos_yaw * world["world_x"] - sin_yaw * world["world_z"]
up = world["world_y"]
front = -sin_yaw * world["world_x"] - cos_yaw * world["world_z"]
return np.stack([w, -right, up, front]).astype(np.float32)
def decay_fit(signal: np.ndarray, sample_rate: int, upper_db: float, lower_db: float):
energy = np.square(signal.astype(np.float64))
peak = int(np.argmax(energy))
decay = np.cumsum(energy[peak:][::-1])[::-1]
if not len(decay) or decay[0] <= 0:
return None
db = 10.0 * np.log10(np.maximum(decay / decay[0], 1e-14))
mask = (db <= upper_db) & (db >= lower_db)
if np.count_nonzero(mask) < 20:
return None
times = np.arange(len(db), dtype=np.float64) / sample_rate
slope, intercept = np.polyfit(times[mask], db[mask], 1)
if slope >= 0:
return None
return {"rt60_s": float(-60.0 / slope), "slope_db_per_s": float(slope),
"fit_intercept_db": float(intercept), "fit_points": int(np.count_nonzero(mask))}
def acoustic_metrics(rir_ambix: np.ndarray, sample_rate: int) -> dict:
w = rir_ambix[0].astype(np.float64)
peak = int(np.argmax(np.square(w)))
direct_end = min(len(w), peak + max(1, round(0.003 * sample_rate)))
c50_end = min(len(w), peak + round(0.050 * sample_rate))
eps = 1e-18
direct = float(np.sum(np.square(w[peak:direct_end]))) + eps
reverberant = float(np.sum(np.square(w[direct_end:]))) + eps
early = float(np.sum(np.square(w[peak:c50_end]))) + eps
late = float(np.sum(np.square(w[c50_end:]))) + eps
broadband = {
"first_peak_s": float(peak / sample_rate),
"drr_db": float(10 * math.log10(direct / reverberant)),
"c50_db": float(10 * math.log10(early / late)),
"edt_s": None,
"t20_rt60_s": None,
"t30_rt60_s": None,
}
edt = decay_fit(w, sample_rate, 0.0, -10.0)
t20 = decay_fit(w, sample_rate, -5.0, -25.0)
t30 = decay_fit(w, sample_rate, -5.0, -35.0)
broadband["edt_s"] = None if edt is None else edt["rt60_s"]
broadband["t20_rt60_s"] = None if t20 is None else t20["rt60_s"]
broadband["t30_rt60_s"] = None if t30 is None else t30["rt60_s"]
bands = {}
nyquist = sample_rate / 2.0
for centre in (500, 1000, 2000):
low, high = centre / math.sqrt(2.0), centre * math.sqrt(2.0)
sos = butter(4, [low / nyquist, high / nyquist], btype="bandpass", output="sos")
filtered = sosfilt(sos, w)
fit = decay_fit(filtered, sample_rate, -5.0, -25.0)
bands[str(centre)] = None if fit is None else fit["rt60_s"]
valid_bands = [value for value in bands.values() if value is not None]
return {"broadband": broadband, "t20_rt60_octave_s": bands,
"acceptance_rt60_median_s": None if not valid_bands else float(np.median(valid_bands))}
def main() -> None:
root = Path(__file__).resolve().parents[1]
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--scenarios", type=Path, default=root.parent / "simulation_scenarios" / "pilot_scenarios.json")
parser.add_argument("--scenario-id", default="EASE-V1-01")
parser.add_argument("--scene-dir", type=Path, default=root / "scene" / "static_v1" / "exhibition_hall_base")
parser.add_argument("--materials", type=Path, default=root / "materials" / "static_v1_absorptive.json")
parser.add_argument("--output-dir", type=Path, default=root / "assets" / "static_v1" / "rir" / "EASE-V1-01")
parser.add_argument("--target-rt60-s", type=float, nargs=2,
metavar=("MIN", "MAX"), help="override the scenario-wide acceptance interval")
parser.add_argument("--rays", type=int, default=15000)
parser.add_argument("--threads", type=int, default=8)
args = parser.parse_args()
scenario_doc = json.loads(args.scenarios.read_text(encoding="utf-8"))
scenario = next(item for item in scenario_doc["scenarios"] if item["id"] == args.scenario_id)
renderer = dict(DEFAULT_RENDERER)
renderer["sampleRate"] = int(scenario_doc["audio_contract"]["sample_rate_hz"])
renderer["indirectRayCount"] = args.rays
renderer["sourceRayCount"] = max(500, args.rays // 10)
renderer["threadCount"] = args.threads
layout_path = (args.scene_dir / "layout.json").resolve()
if not layout_path.is_file():
raise FileNotFoundError(layout_path)
layout = json.loads(layout_path.read_text(encoding="utf-8"))
stage_config = Path(layout["stage_config"]).resolve()
dataset_config = Path(layout["scene_dataset_config"]).resolve()
semantic_glb = Path(layout["semantic_asset"]).resolve()
materials = args.materials.resolve()
for required in (stage_config, dataset_config, semantic_glb, materials, args.scenarios):
if not required.is_file():
raise FileNotFoundError(required)
output_dir = args.output_dir.resolve()
output_dir.mkdir(parents=True, exist_ok=True)
listener_position = scenario["scene_sampling"]["receiver_position_candidates"][0]
yaw_values = scenario["scene_sampling"]["receiver_yaw_deg_values"]
listener_yaw = float(min(yaw_values, key=lambda value: abs(float(value))))
sim = create_simulator(stage_config, dataset_config, materials, renderer)
raw_rirs = {}
try:
set_listener_identity(sim, listener_position)
for name, (position, _) in CALIBRATION_SOURCES.items():
print(f"render calibration {name} @ {position}", flush=True)
raw_rirs[name] = render(sim, position)
mapping = infer_native_foa(raw_rirs)
print("inferred native FOA mapping: " + json.dumps(mapping, ensure_ascii=False), flush=True)
records = []
for source in scenario["sources"]:
print(f"render {scenario['id']} source {source['id']} @ {source['position']}", flush=True)
raw = render(sim, source["position"])
canonical = native_to_ambix_robot(raw, mapping, listener_yaw)
raw_path = output_dir / f"{source['id']}.native_n3d.npy"
canonical_path = output_dir / f"{source['id']}.ambix_sn3d.npy"
np.save(raw_path, raw, allow_pickle=False)
np.save(canonical_path, canonical, allow_pickle=False)
metrics = acoustic_metrics(canonical, renderer["sampleRate"])
records.append({
"source_id": source["id"], "source_class": source["source_class"],
"source_position_xyz_m": source["position"],
"nominal_azimuth_deg": source["nominal_azimuth_deg"],
"native_rir": raw_path.name, "canonical_rir": canonical_path.name,
"shape": list(canonical.shape), "metrics": metrics,
})
finally:
sim.close()
target_rt60 = args.target_rt60_s or scenario["acoustic_profile"]["target_rt60_s_range"]
medians = [item["metrics"]["acceptance_rt60_median_s"] for item in records]
acceptance = all(value is not None and target_rt60[0] <= value <= target_rt60[1] for value in medians)
fingerprint_payload = {
"scenario": scenario, "renderer": renderer,
"stage_sha256": file_hash(stage_config), "semantic_glb_sha256": file_hash(semantic_glb),
"materials_sha256": file_hash(materials), "listener_position": listener_position, "listener_yaw_deg": listener_yaw,
}
manifest = {
"renderer": "SoundSpaces 2.0 / Habitat-Sim AudioSensor",
"scenario_id": scenario["id"], "sample_rate_hz": renderer["sampleRate"],
"receiver_position_xyz_m": listener_position,
"receiver_yaw_deg": listener_yaw,
"scene": {
"geometry_id": layout["geometry_id"],
"room_size_variant": layout.get("room_size_variant", "base"),
"room_size_m": layout["room_size_m"],
"layout_template": layout.get("layout_template", "open_center"),
"center_occluder_type": layout.get("center_occluder_type", "none"),
"layout_sha256": file_hash(layout_path),
},
"materials_file": str(materials),
"materials_sha256": file_hash(materials),
"output_format": "AmbiX ACN SN3D [W,Y,Z,X], robot-relative",
"native_format": "SoundSpaces four-channel Ambisonics N3D; mapping inferred, not assumed",
"materials_enabled": True, "semantic_mesh_required_and_loaded": True,
"renderer_config": renderer, "native_foa_calibration": mapping,
"target_rt60_s_range": target_rt60, "rt60_acceptance_passed": acceptance,
"acoustic_config_sha256": canonical_json_hash(fingerprint_payload), "rirs": records,
}
manifest_path = output_dir / "rir_manifest.json"
manifest_path.write_text(json.dumps(manifest, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
print(f"wrote {manifest_path}", flush=True)
print(f"RT60 acceptance: {acceptance}; medians={medians}; target={target_rt60}", flush=True)
if not acceptance:
raise SystemExit(2)
if __name__ == "__main__":
main()
|