File size: 34,376 Bytes
21f2aa3 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 | #!/usr/bin/env python3
"""
TajweedSST - Step 3: Physics & Signal Processing Validator
Validates Tajweed rules using acoustic signal analysis:
- Qalqalah: RMS energy dip→spike pattern
- Madd: Duration vs Rate of Speech ratio
- Ghunnah: Formant analysis + nasalization detection
- Tafkheem: F2 formant depression
"""
import numpy as np
from dataclasses import dataclass, field
from typing import List, Dict, Optional, Tuple
from enum import Enum
# Import signal processing libraries
try:
import librosa
HAS_LIBROSA = True
except ImportError:
HAS_LIBROSA = False
print("Warning: librosa not installed. RMS/ZCR analysis unavailable.")
try:
import parselmouth
from parselmouth.praat import call
HAS_PARSELMOUTH = True
except ImportError:
HAS_PARSELMOUTH = False
print("Warning: parselmouth not installed. Formant analysis unavailable.")
class ValidationStatus(Enum):
PASS = "PASS"
FAIL = "FAIL"
MARGINAL = "MARGINAL"
SKIPPED = "SKIPPED"
@dataclass
class PhysicsResult:
"""Result of a physics/signal analysis check"""
status: ValidationStatus
metric_name: str
expected_pattern: str
observed_pattern: str
score: float # 0.0 to 1.0
details: Dict = field(default_factory=dict)
@dataclass
class QalqalahResult(PhysicsResult):
"""Specific result for Qalqalah check"""
rms_profile: str = "" # "dip_then_spike", "flat", "spike_only"
dip_depth: float = 0.0
spike_height: float = 0.0
closure_duration_ms: float = 0.0
@dataclass
class MaddResult(PhysicsResult):
"""Specific result for Madd elongation check"""
actual_duration_ms: float = 0.0
expected_duration_ms: float = 0.0
ratio: float = 0.0 # Actual / Average vowel
@dataclass
class GhunnahResult(PhysicsResult):
"""Specific result for Ghunnah nasalization check"""
nasal_formant_detected: bool = False
pitch_stability: float = 0.0
duration_elongation: float = 0.0
@dataclass
class TafkheemResult(PhysicsResult):
"""Specific result for Tafkheem check"""
f2_value_hz: float = 0.0
f2_baseline_hz: float = 1500.0 # Average F2 for light sounds
depression_ratio: float = 0.0
class PhysicsValidator:
"""
Validates Tajweed rules using signal processing
"""
# Thresholds for validation - tuned for real Abdul Basit recitation
QALQALAH_DIP_THRESHOLD = 0.08 # RMS must drop by 8%
QALQALAH_SPIKE_THRESHOLD = 0.15 # RMS must rise by 15%
MADD_RATIO_ASLI = 1.0 # 1.0x average vowel (baseline)
MADD_RATIO_WAJIB = 2.0 # 2.0x average vowel
MADD_RATIO_LAZIM = 3.5 # 3.5x average vowel
GHUNNAH_MIN_DURATION_MS = 30.0 # Very relaxed
TAFKHEEM_F2_MAX_HZ = 1500.0 # Maximum tolerance for F2
VALIDATION_TOLERANCE = 0.4 # 40% tolerance for all validations
# Precision thresholds - tuned for Arabic letters which can be very short
MIN_SEGMENT_MS = 30.0 # Minimum segment duration for valid analysis
MIN_SEGMENT_SAMPLES = 661 # ~30ms at 22050 Hz
def __init__(self, sample_rate: int = 22050):
self.sample_rate = sample_rate
self._audio_cache = {}
self._average_vowel_duration = 0.1 # Will be calibrated per reciter
def load_audio(self, audio_path: str) -> np.ndarray:
"""Load audio file, with caching"""
if audio_path not in self._audio_cache:
if HAS_LIBROSA:
y, sr = librosa.load(audio_path, sr=self.sample_rate)
self._audio_cache[audio_path] = y
else:
# Fallback: generate noise for testing
self._audio_cache[audio_path] = np.random.randn(self.sample_rate * 10) * 0.1
return self._audio_cache[audio_path]
def safe_extract_segment(self, audio: np.ndarray, start: float, end: float) -> tuple:
"""
PRECISION: Safely extract audio segment with bounds and validity checking.
Returns:
tuple: (segment, is_valid, error_reason)
"""
# Bounds checking
start_sample = max(0, int(start * self.sample_rate))
end_sample = min(len(audio), int(end * self.sample_rate))
# Sanity check
if start_sample >= end_sample:
return None, False, "invalid_range"
segment = audio[start_sample:end_sample]
# Length check
if len(segment) < self.MIN_SEGMENT_SAMPLES:
return segment, False, f"too_short_{len(segment)}_samples"
# NaN/Inf check
if np.any(np.isnan(segment)) or np.any(np.isinf(segment)):
segment = np.nan_to_num(segment, nan=0.0, posinf=0.0, neginf=0.0)
return segment, True, None
def safe_rms(self, segment: np.ndarray, frame_length: int = 256, hop_length: int = 64) -> np.ndarray:
"""
PRECISION: Calculate RMS with NaN protection.
"""
if not HAS_LIBROSA:
return np.array([0.0])
rms = librosa.feature.rms(y=segment, frame_length=frame_length, hop_length=hop_length)[0]
# Protect against NaN/Inf
rms = np.nan_to_num(rms, nan=0.0, posinf=1.0, neginf=0.0)
# Normalize to prevent division issues
if np.max(rms) > 0:
rms = rms / np.max(rms)
return rms
def validate_qalqalah(self,
audio: np.ndarray,
start: float,
end: float) -> QalqalahResult:
"""
Validate Qalqalah rule: Must show closure (RMS dip) then release (RMS spike)
Physics: The "bounce" is caused by complete oral closure followed by
abrupt release. RMS energy shows: stable→dip→spike pattern.
"""
if not HAS_LIBROSA:
return QalqalahResult(
status=ValidationStatus.SKIPPED,
metric_name="RMS Energy",
expected_pattern="dip_then_spike",
observed_pattern="unknown",
score=0.0,
rms_profile="unknown"
)
# PRECISION: Use safe extraction
segment, is_valid, error = self.safe_extract_segment(audio, start, end)
if not is_valid:
return QalqalahResult(
status=ValidationStatus.SKIPPED,
metric_name="RMS Energy",
expected_pattern="dip_then_spike",
observed_pattern=error or "invalid_segment",
score=0.0,
rms_profile="unknown",
details={"reason": error}
)
# PRECISION: Use safe RMS with NaN protection
rms = self.safe_rms(segment)
if len(rms) < 3:
return QalqalahResult(
status=ValidationStatus.SKIPPED,
metric_name="RMS Energy",
expected_pattern="dip_then_spike",
observed_pattern="insufficient_frames",
score=0.0,
rms_profile="unknown",
details={"reason": f"Only {len(rms)} RMS frames < 3 minimum"}
)
# Analyze RMS pattern
# Qalqalah should show: high → dip → spike
# Find minimum and maximum in second half (release)
midpoint = len(rms) // 2
# First half: Find the dip (closure)
first_half_mean = np.mean(rms[:midpoint]) if midpoint > 0 else rms[0]
dip_idx = np.argmin(rms)
dip_value = rms[dip_idx]
# Second half: Find the spike (release)
spike_idx = midpoint + np.argmax(rms[midpoint:]) if midpoint < len(rms) else len(rms) - 1
spike_value = rms[spike_idx] if spike_idx < len(rms) else rms[-1]
# Calculate metrics
dip_depth = (first_half_mean - dip_value) / first_half_mean if first_half_mean > 0 else 0
spike_height = (spike_value - dip_value) / dip_value if dip_value > 0 else 0
# Determine pattern
if dip_depth >= self.QALQALAH_DIP_THRESHOLD and spike_height >= self.QALQALAH_SPIKE_THRESHOLD:
rms_profile = "dip_then_spike"
status = ValidationStatus.PASS
score = min(1.0, (dip_depth + spike_height) / 2)
elif spike_height >= self.QALQALAH_SPIKE_THRESHOLD:
rms_profile = "spike_only"
status = ValidationStatus.MARGINAL
score = spike_height / 2
else:
rms_profile = "flat"
status = ValidationStatus.FAIL
score = 0.0
# Estimate closure duration (using safe_rms default hop_length=64)
if dip_idx > 0:
frames_to_dip = dip_idx
closure_duration_ms = (frames_to_dip * 64 / self.sample_rate) * 1000
else:
closure_duration_ms = 0.0
return QalqalahResult(
status=status,
metric_name="RMS Energy",
expected_pattern="dip_then_spike",
observed_pattern=rms_profile,
score=score,
rms_profile=rms_profile,
dip_depth=dip_depth,
spike_height=spike_height,
closure_duration_ms=closure_duration_ms
)
def validate_madd(self,
audio: np.ndarray,
start: float,
end: float,
expected_count: int = 2) -> MaddResult:
"""
Validate Madd rule: Duration must match expected elongation count
Physics: Madd is pure duration comparison.
- Asli (natural): 2 counts
- Wajib (obligatory): 4-5 counts
- Lazim (required): 6 counts
"""
actual_duration = end - start
actual_duration_ms = actual_duration * 1000
# Expected duration based on average vowel and count
expected_duration = self._average_vowel_duration * expected_count
expected_duration_ms = expected_duration * 1000
# Calculate ratio
ratio = actual_duration / self._average_vowel_duration if self._average_vowel_duration > 0 else 0
# Determine pass/fail based on expected count
tolerance = 0.3 # 30% tolerance
if expected_count == 2:
threshold = self.MADD_RATIO_ASLI
elif expected_count == 4:
threshold = self.MADD_RATIO_WAJIB
else:
threshold = self.MADD_RATIO_LAZIM
if ratio >= threshold * (1 - tolerance):
if ratio <= threshold * (1 + tolerance):
status = ValidationStatus.PASS
score = 1.0
else:
status = ValidationStatus.MARGINAL # Too long, but acceptable
score = 0.7
else:
status = ValidationStatus.FAIL
score = ratio / threshold if threshold > 0 else 0
return MaddResult(
status=status,
metric_name="Duration Ratio",
expected_pattern=f"{expected_count}x average vowel",
observed_pattern=f"{ratio:.1f}x average vowel",
score=score,
actual_duration_ms=actual_duration_ms,
expected_duration_ms=expected_duration_ms,
ratio=ratio
)
def validate_ghunnah(self,
audio: np.ndarray,
start: float,
end: float) -> GhunnahResult:
"""
Validate Ghunnah (nasalization) rule
Physics:
- Drop in high-frequency energy (nasal anti-formant ~500Hz)
- Stable pitch during nasalization
- Duration elongation (2 counts minimum)
"""
if not HAS_PARSELMOUTH:
return GhunnahResult(
status=ValidationStatus.SKIPPED,
metric_name="Formant Analysis",
expected_pattern="nasal_resonance",
observed_pattern="unknown",
score=0.0
)
duration_ms = (end - start) * 1000
# Check minimum duration
if duration_ms < self.GHUNNAH_MIN_DURATION_MS:
return GhunnahResult(
status=ValidationStatus.MARGINAL, # PRECISION: Changed from FAIL to MARGINAL
metric_name="Formant Analysis",
expected_pattern="nasal_resonance",
observed_pattern="short_but_valid",
score=duration_ms / self.GHUNNAH_MIN_DURATION_MS,
duration_elongation=duration_ms / self.GHUNNAH_MIN_DURATION_MS,
details={"reason": f"Duration {duration_ms:.1f}ms < {self.GHUNNAH_MIN_DURATION_MS}ms minimum"}
)
# PRECISION: Use safe extraction
segment, is_valid, error = self.safe_extract_segment(audio, start, end)
if not is_valid:
return GhunnahResult(
status=ValidationStatus.SKIPPED,
metric_name="Formant Analysis",
expected_pattern="nasal_resonance",
observed_pattern=error or "invalid_segment",
score=0.0,
details={"reason": error}
)
# Convert to Praat Sound object
try:
import tempfile
import soundfile as sf
with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as f:
sf.write(f.name, segment, self.sample_rate)
sound = parselmouth.Sound(f.name)
# Get pitch for stability analysis
pitch = call(sound, "To Pitch", 0.0, 75, 600)
pitch_values = pitch.selected_array['frequency']
pitch_values = pitch_values[pitch_values > 0] # Remove unvoiced
if len(pitch_values) > 1:
pitch_stability = 1.0 - (np.std(pitch_values) / np.mean(pitch_values))
else:
pitch_stability = 0.0
# Formant analysis for nasal detection
formant = call(sound, "To Formant (burg)", 0.0, 5, 5500, 0.025, 50)
# Nasalization shows anti-resonance around F1 region
# Check for characteristic nasal formant pattern
nasal_formant_detected = True # Simplified detection
except Exception as e:
print(f"Parselmouth error: {e}")
return GhunnahResult(
status=ValidationStatus.SKIPPED,
metric_name="Formant Analysis",
expected_pattern="nasal_resonance",
observed_pattern="analysis_error",
score=0.0
)
# Scoring
duration_score = min(1.0, duration_ms / (self.GHUNNAH_MIN_DURATION_MS * 2))
pitch_score = max(0.0, pitch_stability)
total_score = (duration_score + pitch_score) / 2
if total_score >= 0.7:
status = ValidationStatus.PASS
elif total_score >= 0.4:
status = ValidationStatus.MARGINAL
else:
status = ValidationStatus.FAIL
return GhunnahResult(
status=status,
metric_name="Formant Analysis",
expected_pattern="nasal_resonance",
observed_pattern="analyzed",
score=total_score,
nasal_formant_detected=nasal_formant_detected,
pitch_stability=pitch_stability,
duration_elongation=duration_ms / self.GHUNNAH_MIN_DURATION_MS
)
def validate_tafkheem(self,
audio: np.ndarray,
start: float,
end: float) -> TafkheemResult:
"""
Validate Tafkheem (heavy letter) rule
Physics: Heavy letters show depressed F2 formant
- Normal letters: F2 ~1500 Hz
- Heavy letters: F2 ~1000-1200 Hz
"""
if not HAS_PARSELMOUTH:
return TafkheemResult(
status=ValidationStatus.SKIPPED,
metric_name="F2 Formant",
expected_pattern="F2 < 1200 Hz",
observed_pattern="unknown",
score=0.0
)
# PRECISION: Use safe extraction
segment, is_valid, error = self.safe_extract_segment(audio, start, end)
if not is_valid:
return TafkheemResult(
status=ValidationStatus.SKIPPED,
metric_name="F2 Formant",
expected_pattern=f"F2 < {self.TAFKHEEM_F2_MAX_HZ} Hz",
observed_pattern=error or "invalid_segment",
score=0.0,
details={"reason": error}
)
try:
import tempfile
import soundfile as sf
with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as f:
sf.write(f.name, segment, self.sample_rate)
sound = parselmouth.Sound(f.name)
# Get F2 formant
formant = call(sound, "To Formant (burg)", 0.0, 5, 5500, 0.025, 50)
# Get average F2
f2_values = []
num_frames = call(formant, "Get number of frames")
for i in range(1, num_frames + 1):
f2 = call(formant, "Get value at time", 2, call(formant, "Get time from frame number", i), "Hertz", "Linear")
if not np.isnan(f2) and f2 > 0:
f2_values.append(f2)
if f2_values:
f2_mean = np.mean(f2_values)
else:
f2_mean = 0
except Exception as e:
print(f"Parselmouth error: {e}")
return TafkheemResult(
status=ValidationStatus.SKIPPED,
metric_name="F2 Formant",
expected_pattern="F2 < 1200 Hz",
observed_pattern="analysis_error",
score=0.0
)
# Calculate depression ratio
baseline_f2 = 1500.0
depression_ratio = (baseline_f2 - f2_mean) / baseline_f2 if f2_mean > 0 and f2_mean < baseline_f2 else 0
# Scoring
if f2_mean <= self.TAFKHEEM_F2_MAX_HZ:
status = ValidationStatus.PASS
score = 1.0
elif f2_mean <= 1350:
status = ValidationStatus.MARGINAL
score = 0.6
else:
status = ValidationStatus.FAIL
score = max(0.0, depression_ratio)
return TafkheemResult(
status=status,
metric_name="F2 Formant",
expected_pattern=f"F2 < {self.TAFKHEEM_F2_MAX_HZ} Hz",
observed_pattern=f"F2 = {f2_mean:.0f} Hz",
score=score,
f2_value_hz=f2_mean,
f2_baseline_hz=baseline_f2,
depression_ratio=depression_ratio
)
# =========================================================================
# NEW VALIDATORS: Complete Tajweed Physics Coverage
# =========================================================================
def validate_idgham(self,
audio: np.ndarray,
nun_start: float,
nun_end: float,
next_letter_end: float,
has_ghunnah: bool = True) -> PhysicsResult:
"""
Validate Idgham (assimilation) rule
Physics:
- Full Idgham (ر/ل): Complete merger, smooth energy, no nun boundary
- Partial Idgham (ي/ن/م/و): Ghunnah preserved during transition
"""
if not HAS_LIBROSA:
return PhysicsResult(
status=ValidationStatus.SKIPPED,
metric_name="Energy Continuity",
expected_pattern="smooth_transition",
observed_pattern="unknown",
score=0.0
)
# Extract the transition window (nun end to next letter)
start_sample = int(nun_start * self.sample_rate)
end_sample = int(next_letter_end * self.sample_rate)
segment = audio[start_sample:end_sample]
if len(segment) < 100:
return PhysicsResult(
status=ValidationStatus.FAIL,
metric_name="Energy Continuity",
expected_pattern="smooth_transition",
observed_pattern="segment_too_short",
score=0.0
)
# Calculate RMS to check for smooth energy transition
frame_length = 256
hop_length = 64
rms = librosa.feature.rms(y=segment, frame_length=frame_length, hop_length=hop_length)[0]
# Calculate energy variance - low variance = smooth transition
rms_variance = np.std(rms) / np.mean(rms) if np.mean(rms) > 0 else 1.0
# For Idgham, we expect smooth continuous energy (low variance)
smoothness_score = 1.0 - min(1.0, rms_variance)
# Check for boundary sharpness (should be LOW for Idgham)
rms_diff = np.abs(np.diff(rms))
max_jump = np.max(rms_diff) / np.mean(rms) if np.mean(rms) > 0 else 0
boundary_score = 1.0 - min(1.0, max_jump)
total_score = (smoothness_score + boundary_score) / 2
if total_score >= 0.6:
status = ValidationStatus.PASS
elif total_score >= 0.4:
status = ValidationStatus.MARGINAL
else:
status = ValidationStatus.FAIL
return PhysicsResult(
status=status,
metric_name="Energy Continuity",
expected_pattern="smooth_transition" if not has_ghunnah else "smooth_with_ghunnah",
observed_pattern=f"smoothness={smoothness_score:.2f}",
score=total_score,
details={"smoothness": smoothness_score, "boundary_score": boundary_score}
)
def validate_ikhfa(self,
audio: np.ndarray,
start: float,
end: float) -> PhysicsResult:
"""
Validate Ikhfa (concealment) rule
Physics:
- Gradual nasalization transition (not abrupt like pure Ghunnah)
- Partial nasal resonance that fades
"""
if not HAS_LIBROSA:
return PhysicsResult(
status=ValidationStatus.SKIPPED,
metric_name="Nasalization Gradient",
expected_pattern="gradual_nasal",
observed_pattern="unknown",
score=0.0
)
start_sample = int(start * self.sample_rate)
end_sample = int(end * self.sample_rate)
segment = audio[start_sample:end_sample]
if len(segment) < 100:
return PhysicsResult(
status=ValidationStatus.FAIL,
metric_name="Nasalization Gradient",
expected_pattern="gradual_nasal",
observed_pattern="segment_too_short",
score=0.0
)
# Split into thirds to check for gradient
third = len(segment) // 3
# Calculate spectral centroid (nasal sounds have lower centroid)
sc = librosa.feature.spectral_centroid(y=segment, sr=self.sample_rate)[0]
if len(sc) < 3:
return PhysicsResult(
status=ValidationStatus.FAIL,
metric_name="Nasalization Gradient",
expected_pattern="gradual_nasal",
observed_pattern="insufficient_frames",
score=0.0
)
# Check for gradient pattern: centroid should change gradually
sc_diff = np.abs(np.diff(sc))
gradient_smoothness = 1.0 - min(1.0, np.std(sc_diff) / np.mean(sc_diff)) if np.mean(sc_diff) > 0 else 0.5
# Duration check (Ikhfa should have reasonable duration)
duration_ms = (end - start) * 1000
duration_score = min(1.0, duration_ms / 100) if duration_ms > 0 else 0
total_score = (gradient_smoothness + duration_score) / 2
if total_score >= 0.6:
status = ValidationStatus.PASS
elif total_score >= 0.4:
status = ValidationStatus.MARGINAL
else:
status = ValidationStatus.FAIL
return PhysicsResult(
status=status,
metric_name="Nasalization Gradient",
expected_pattern="gradual_nasal",
observed_pattern=f"gradient={gradient_smoothness:.2f}",
score=total_score,
details={"gradient_smoothness": gradient_smoothness, "duration_ms": duration_ms}
)
def validate_iqlab(self,
audio: np.ndarray,
start: float,
end: float) -> PhysicsResult:
"""
Validate Iqlab (ن→م before ب)
Physics:
- Same as Ghunnah but with bilabial closure
- Nasal formant + lip closure pattern (F1/F2 characteristic of /m/)
"""
# Iqlab is essentially Ghunnah with bilabial characteristics
# Reuse ghunnah validation logic
ghunnah_result = self.validate_ghunnah(audio, start, end)
# Modify result type for Iqlab
return PhysicsResult(
status=ghunnah_result.status,
metric_name="Bilabial Nasal",
expected_pattern="mim_like_nasal",
observed_pattern=ghunnah_result.observed_pattern,
score=ghunnah_result.score,
details={"ghunnah_check": ghunnah_result.status.value}
)
def validate_izhar(self,
audio: np.ndarray,
letter_start: float,
letter_end: float,
next_letter_start: float) -> PhysicsResult:
"""
Validate Izhar (clear pronunciation)
Physics:
- Clean, sharp boundary between letters
- No nasalization
- Clear articulation energy pattern
"""
if not HAS_LIBROSA:
return PhysicsResult(
status=ValidationStatus.SKIPPED,
metric_name="Boundary Sharpness",
expected_pattern="clean_boundary",
observed_pattern="unknown",
score=0.0
)
# Check boundary region
boundary_start = max(0, letter_end - 0.02) # 20ms before boundary
boundary_end = min(len(audio) / self.sample_rate, next_letter_start + 0.02) # 20ms after
start_sample = int(boundary_start * self.sample_rate)
end_sample = int(boundary_end * self.sample_rate)
segment = audio[start_sample:end_sample]
if len(segment) < 50:
return PhysicsResult(
status=ValidationStatus.FAIL,
metric_name="Boundary Sharpness",
expected_pattern="clean_boundary",
observed_pattern="segment_too_short",
score=0.0
)
# Calculate RMS to find sharp transitions
frame_length = 128
hop_length = 32
rms = librosa.feature.rms(y=segment, frame_length=frame_length, hop_length=hop_length)[0]
# Look for clear dip/change at boundary
rms_diff = np.abs(np.diff(rms))
max_change = np.max(rms_diff) / np.mean(rms) if np.mean(rms) > 0 else 0
# High change = sharp boundary = good for Izhar
sharpness_score = min(1.0, max_change)
if sharpness_score >= 0.3: # Clear boundary detected
status = ValidationStatus.PASS
score = min(1.0, sharpness_score * 2)
elif sharpness_score >= 0.15:
status = ValidationStatus.MARGINAL
score = sharpness_score * 2
else:
status = ValidationStatus.FAIL
score = sharpness_score
return PhysicsResult(
status=status,
metric_name="Boundary Sharpness",
expected_pattern="clean_boundary",
observed_pattern=f"sharpness={sharpness_score:.2f}",
score=score,
details={"boundary_sharpness": sharpness_score}
)
def validate_tarqeeq(self,
audio: np.ndarray,
start: float,
end: float) -> PhysicsResult:
"""
Validate Tarqeeq (light letters) - opposite of Tafkheem
Physics: Light letters show elevated F2 formant (F2 > 1400 Hz)
"""
# Reuse Tafkheem logic but invert the threshold
tafkheem_result = self.validate_tafkheem(audio, start, end)
if tafkheem_result.status == ValidationStatus.SKIPPED:
return PhysicsResult(
status=ValidationStatus.SKIPPED,
metric_name="F2 Formant",
expected_pattern="F2 > 1400 Hz",
observed_pattern="unknown",
score=0.0
)
# For Tarqeeq, we want HIGH F2 (opposite of Tafkheem)
f2_value = tafkheem_result.details.get('f2_value_hz', tafkheem_result.f2_value_hz if hasattr(tafkheem_result, 'f2_value_hz') else 0)
TARQEEQ_F2_MIN_HZ = 1400.0
if f2_value >= TARQEEQ_F2_MIN_HZ:
status = ValidationStatus.PASS
score = 1.0
elif f2_value >= 1300:
status = ValidationStatus.MARGINAL
score = 0.6
else:
status = ValidationStatus.FAIL
score = f2_value / TARQEEQ_F2_MIN_HZ if f2_value > 0 else 0
return PhysicsResult(
status=status,
metric_name="F2 Formant",
expected_pattern=f"F2 > {TARQEEQ_F2_MIN_HZ} Hz",
observed_pattern=f"F2 = {f2_value:.0f} Hz",
score=score,
details={"f2_value_hz": f2_value}
)
def validate_sakt(self,
audio: np.ndarray,
start: float,
end: float) -> PhysicsResult:
"""
Validate Sakt (brief pause without breath)
Physics:
- Brief silence (50-200ms)
- RMS below threshold
- No breathing artifacts
"""
if not HAS_LIBROSA:
return PhysicsResult(
status=ValidationStatus.SKIPPED,
metric_name="Silence Detection",
expected_pattern="brief_silence",
observed_pattern="unknown",
score=0.0
)
start_sample = int(start * self.sample_rate)
end_sample = int(end * self.sample_rate)
segment = audio[start_sample:end_sample]
duration_ms = (end - start) * 1000
if len(segment) < 10:
return PhysicsResult(
status=ValidationStatus.FAIL,
metric_name="Silence Detection",
expected_pattern="brief_silence",
observed_pattern="segment_too_short",
score=0.0
)
# Calculate RMS
rms = np.sqrt(np.mean(segment**2))
# Thresholds
SAKT_RMS_THRESHOLD = 0.05
SAKT_MIN_MS = 50
SAKT_MAX_MS = 200
# Check RMS (should be very low)
is_silent = rms < SAKT_RMS_THRESHOLD
# Check duration
duration_ok = SAKT_MIN_MS <= duration_ms <= SAKT_MAX_MS
if is_silent and duration_ok:
status = ValidationStatus.PASS
score = 1.0
elif is_silent and (duration_ms > 30):
status = ValidationStatus.MARGINAL
score = 0.6
else:
status = ValidationStatus.FAIL
score = 0.0 if rms >= SAKT_RMS_THRESHOLD else 0.3
return PhysicsResult(
status=status,
metric_name="Silence Detection",
expected_pattern=f"silence_{SAKT_MIN_MS}-{SAKT_MAX_MS}ms",
observed_pattern=f"rms={rms:.3f}, dur={duration_ms:.0f}ms",
score=score,
details={"rms": rms, "duration_ms": duration_ms, "is_silent": is_silent}
)
def calibrate_average_vowel(self, audio: np.ndarray, vowel_segments: List[Tuple[float, float]]) -> float:
"""
Calibrate average vowel duration for this reciter
This is crucial for Madd validation as reciter pace varies
"""
if not vowel_segments:
return 0.1 # Default 100ms
durations = [end - start for start, end in vowel_segments]
self._average_vowel_duration = np.mean(durations)
return self._average_vowel_duration
def main():
"""Test physics validator"""
print("=" * 50)
print("TajweedSST Physics Validator Test")
print("=" * 50)
# Create mock audio
sample_rate = 22050
duration = 2.0
t = np.linspace(0, duration, int(sample_rate * duration))
# Create a test signal with dip→spike pattern (simulating Qalqalah)
audio = np.sin(2 * np.pi * 440 * t) * 0.5
# Add dip in middle
dip_start = int(len(audio) * 0.4)
dip_end = int(len(audio) * 0.5)
audio[dip_start:dip_end] *= 0.1
# Add spike after dip
spike_start = int(len(audio) * 0.5)
spike_end = int(len(audio) * 0.6)
audio[spike_start:spike_end] *= 2.0
validator = PhysicsValidator(sample_rate=sample_rate)
# Test Qalqalah
print("\nQalqalah Test:")
result = validator.validate_qalqalah(audio, 0.3, 0.8)
print(f" Status: {result.status.value}")
print(f" Profile: {result.rms_profile}")
print(f" Score: {result.score:.2f}")
print(f" Dip Depth: {result.dip_depth:.2f}")
print(f" Spike Height: {result.spike_height:.2f}")
# Test Madd
print("\nMadd Test:")
validator._average_vowel_duration = 0.1 # 100ms average
result = validator.validate_madd(audio, 0.0, 0.4, expected_count=4)
print(f" Status: {result.status.value}")
print(f" Ratio: {result.ratio:.1f}x")
print(f" Score: {result.score:.2f}")
if __name__ == "__main__":
main()
|