- attempt to make MOSes better

bin/aqua.cfg

@@ -0,0 +1,21 @@
+AQuA:
+    mode:       files
+    # src:      file test_audio/jane_8k.wav
+    # tstf:     test_audio/jane_8k_40.wav
+    avlp:       off
+    smtnrm:     off
+    decor:      off
+    mprio:      off
+    acr:        auto
+    npnt:       auto
+    voip:       on
+    enorm:      rms
+    g711:       off
+    spfrcor:    on
+    grad:       off
+    tmc:        on
+    miter:      1
+    ratem:      "%%m"
+    trim:       "r 15"
+    output:     json
+    

bin/macos/README.txt

@@ -1 +0,0 @@
-This pjsua requires 10.14 at least!

bin/pvqa.cfg

@@ -1,226 +1,232 @@
-BOF Common
-  IntervalLength = 0.68
-  IsUseUncertain = false
-  IsUseMixMode = true
-  IsUseDistance = false
-  AllWeight = 1.0
-  SilWeight = 1
-  VoiWeight = 1
-  AllCoefficient = 1.0
-  SilCoefficient = 1.0
-  VoiCoefficient = 1.0
-  SilThreshold = -37.50
-  IsOnePointSil = false
-  IsNormResult = true
-  IsMapScore = true
-EOF Common
+Common:
+    IntervalLength:       0.68
+    IsUseUncertain:       no
+    IsUseMixMode:         yes
+    IsUseDistance:        no
+    AllWeight:            1.0
+    SilWeight:            1.0
+    VoiWeight:            1.0
+    AllCoefficient:       1.0
+    SilCoefficient:       1.0
+    VoiCoefficient:       1.0
+    SilThreshold:         -37.50
+    IsOnePointSil:        no
+    IsNormResult:         yes
+    IsMapScore:           yes
+    NormalizeByRms:       yes
 
-BOF Detector
-  Name = SNR
-  DetectorType = SNR
-  IntThresh = 0.10
-  FrameThresh = 14
-  DetThresh = 0.10
-  PVQA-Flag = true
-  PVQA-Weight = 1.0
-  DetMode = Both
-EOF Detector
+SilenceEraser:
+    Enabled:              no
+    Options:
 
-BOF Detector
-  Name = DeadAir-00
-  DetectorType = DeadAir
-  IntThresh = 0.60
-  DetThresh = 0.60
-  PVQA-Flag = true
-  PVQA-Weight = 1.0
-  DetMode = Both
-EOF Detector
+Detector:
+  - Name:                 SNR
+    DetectorType:         SNR
+    IntThresh:            0.10
+    FrameThresh:          14
+    DetThresh:            0.10
+    PVQA-Flag:            yes
+    PVQA-Weight:          1.0
+    DetMode:              both
 
-BOF Detector
-  Name = DeadAir-01
-  DetectorType = DeadAir
-  IntThresh = 0.5
-  DetThresh = 0.5
-  PVQA-Flag = true
-  PVQA-Weight = 1.0
-  DetMode = Both
-EOF Detector
+  - Name:                 Noise
+    DetectorType:         Noise
+    IntThresh:            0.99
+    DetThresh:            0.99
+    # This is still experimental detector so its values are not participating in MOS calculation
+    PVQA-Flag:            false
+    PVQA-Weight:          1.0
+    DetMode:              both
 
-BOF Detector
-  Name = Click
-  DetectorType = Clicking
-  IntThresh = 0.10
-  DetThresh = 0.10
-  PVQA-Flag = true
-  PVQA-Weight = 1.0
-  DetMode = Both
-EOF Detector
+  - Name:                 DTMF
+    DetectorType:         DTMF
+    IntThresh:            0.99
+    DetThresh:            0.99
 
-BOF Detector
-  Name = VAD-Clipping
-  DetectorType = VADClipping
-  IntThresh = 0.0
-  FrameThresh = 0.0
-  DetThresh = 0.0
-  PVQA-Flag = true
-  PVQA-Weight = 1.0
-  DetMode = Both
-EOF Detector
+    # There is no sense to use detected DTMF signal in MOS calculation in the current config
+    PVQA-Flag:            no
+    PVQA-Weight:          0.0
+    DetMode:              both
 
-BOF Detector
-  Name = Amplitude-Clipping
-  DetectorType = AmpClipping
-  IntThresh = 0.00
-  FrameThresh = 1.00
-  DetThresh = 0.00
-  PVQA-Flag = true
-  PVQA-Weight = 1.00
-  DetMode = Both
-EOF Detector
+  - Name:                 DeadAir-00
+    DetectorType:         DeadAir
+    IntThresh:            0.60
+    DetThresh:            0.60
+    PVQA-Flag:            true
+    PVQA-Weight:          1.0
+    DetMode:              both
 
-BOF Detector
-  Name = Dynamic-Clipping
-  DetectorType = AmpClipping
-  IntThresh = 0.05
-  FrameThresh = 1.50
-  DetThresh = 0
-  PVQA-Flag = true
-  PVQA-Weight = 0.0
-  DetMode = Voice
-EOF Detector
-
-BOF Base EchoMono
-  SamplesType = UnKnownCodec
-  StepLengthSec = 0.5
-  MinDelayMs = 50
-  MaxLengthMs = 2800
-  WindowFunckID = 0
-  SpanLengthMs = 50
-EOF Base EchoMono
-
-BOF Detector
-  Name = ECHO
-  DetectorType = EchoMono
-  IntThresh = 0.00
-  FrameThresh = -40.0
-  DetThresh = 0.00
-  PVQA-Flag = true
-  PVQA-Weight = 1.0
-  DetMode = Voice
-  STAT-Flag = true
-  SpanLengthMs = 50
-EOF Detector
-
-BOF Detector
-  Name = Silent-Call-Detection
-  DetectorType = DeadAir
-  IntThresh = 0.99
-  DetThresh = 0.99
-  PVQA-Flag = false
-  PVQA-Weight = 1.0
-EOF Detector
-
-BOF Base SNR
-  MinPowerThresh = 1.0000
-  LogEnergyCoefficient = 10.0000
-  MinSignalLevel = 40.0000
-  MinSNRDelta = 0.0001
-  MinEnergyDisp = 3.0000
-  MinEnergyDelta = 1.0000
-  SamplesType = UnKnownCodec
-EOF Base SNR
-
-BOF Base AmpClipping
-  FlyAddingCoefficient = 0.1000
-  IsUseDynamicClipping = false
-  SamplesType = UnKnownCodec
-EOF Base AmpClipping
-
-BOF Base Clicking
-  SamplesType = UnKnownCodec
-EOF Base Clicking
-
-BOF Base DeadAir
-  StuckDeltaThreshold = 6
-  MinNonStuckTime = 80
-  MinStuckTime = 80
-  MinStartNonStuckTime = 1920
-  MinLevelThreshold = 256
-  SamplesType = UnKnownCodec
-EOF Base DeadAir
+  - Name:                 DeadAir-01
+    DetectorType:         DeadAir
+    IntThresh:            0.5
+    DetThresh:            0.5
+    PVQA-Flag:            yes
+    PVQA-Weight:          1.0
+    DetMode:              both
+    Override:
+        MinLevelThreshold:    0
 
-BOF Base VADClipping
-  SamplesType = UnKnownCodec
-EOF Base VADClipping
+  - Name:                 Click
+    DetectorType:         Clicking
+    IntThresh:            0.10
+    DetThresh:            0.10
+    PVQA-Flag:            true
+    PVQA-Weight:          1.0
+    DetMode:              both
 
-BOF DeadAir-01
-  MinLevelThreshold = 0
-EOF DeadAir-01
+  - Name:                 VAD-Clipping
+    DetectorType:         VADClipping
+    IntThresh:            0.0
+    FrameThresh:          0.0
+    DetThresh:            0.0
+    PVQA-Flag:            true
+    PVQA-Weight:          1.0
+    DetMode:              both
 
-BOF Silent-Call-Detection
-  MinLevelThreshold = 0
-  IsUseRMSPower = true
-  MinRMSThreshold = -70
-EOF Silent-Call-Detection
+  - Name:                 AmpClipping
+    DetectorType:         AmpClipping
+    IntThresh:            0.00
+    FrameThresh:          1.00
+    DetThresh:            0.00
+    PVQA-Flag:            true
+    PVQA-Weight:          1.00
+    DetMode:              both
 
-BOF Dynamic-Clipping
-  FlyAddingCoefficient = 0.1000
-  SamplesType = UnKnownCodec
-  IsUseDynamicClipping = true
-EOF Dynamic-Clipping
+  - Name:                 DynClipping
+    DetectorType:         AmpClipping
+    IntThresh:            0.05
+    FrameThresh:          1.50
+    DetThresh:            0
+    PVQA-Flag:            true
+    PVQA-Weight:          0.0
+    DetMode:              voice
+    Override:
+        FlyAddingCoefficient: 0.1000
+        SamplesType:          UnKnownCodec
+        IsUseDynamicClipping: yes
 
-BOF Correction
-  IntStart = 5.0
-  IntEnd = 4.2
-  Mult = 1.0
-  #Shift = -1.7
-  Shift = 0
-EOF Correction
+  - Name:                 Echo
+    DetectorType:         EchoMono
+    IntThresh:            0.00
+    FrameThresh:          -40.0
+    DetThresh:            0.00
+    PVQA-Flag:            true
+    PVQA-Weight:          1.0
+    DetMode:              voice
+    STAT-Flag:            true
+    SpanLengthMs:         50
 
-BOF Correction
-  IntStart = 4.2
-  IntEnd = 3.5
-  Mult = 1.0
-  #Shift = -0.85
-  Shift = 0
-EOF Correction
+  - Name:                 SilentCall
+    DetectorType:         DeadAir
+    IntThresh:            0.99
+    DetThresh:            0.99
+    PVQA-Flag:            false
+    PVQA-Weight:          1.0
+    Override:
+        MinLevelThreshold:    0
+        IsUseRMSPower:        yes
+        MinRMSThreshold:      -70
 
-BOF SR Correction
-  SampleRate = 11000.0
-  Shift = 0.05
-EOF SR Correction
 
-BOF SR Correction
-  SampleRate = 16000.0
-  Shift = 0.1
-EOF SR Correction
 
-BOF SR Correction
-  SampleRate = 22000.0
-  Shift = 0.2
-EOF SR Correction
+Base EchoMono:
+  SamplesType:          UnKnownCodec
+  StepLengthSec:        0.5
+  MinDelayMs:           50
+  MaxLengthMs:          2800
+  WindowFunckID:        0
+  SpanLengthMs:         50
 
-BOF SR Correction
-  SampleRate = 32000.0
-  Shift = 0.3
-EOF SR Correction
+Base SNR:
+  MinPowerThresh:       1.0000
+  LogEnergyCoefficient: 10.0000
+  MinSignalLevel:       40.0000
+  MinSNRDelta:          0.0001
+  MinEnergyDisp:        3.0000
+  MinEnergyDelta:       1.0000
+  SamplesType:          UnKnownCodec
 
-BOF SR Correction
-  SampleRate = 48000.0
-  Shift = 0.45
-EOF SR Correction
+Base DTMF:
+  SamplesType:          UnKnownCodec
 
-BOF SR Correction
-  SampleRate = 96000.0
-  Shift = 0.5
-EOF SR Correction
+Base AmpClipping:
+  FlyAddingCoefficient: 0.1000
+  IsUseDynamicClipping: no
+  SamplesType:          UnKnownCodec
 
-BOF SR Correction
-  SampleRate = 192000.0
-  Shift = 0.6
-EOF SR Correction
+Base Clicking:
+  SamplesType:          UnKnownCodec
 
-BOF Scores Map
-  ScoresLine = 4;3.027000;2.935000;2.905000;2.818000;2.590000;2.432000;2.310000;1.665000;1.000000;
-EOF Scores Map
+Base DeadAir:
+  StuckDeltaThreshold:  6
+  MinNonStuckTime:      80
+  MinStuckTime:         80
+  MinStartNonStuckTime: 1920
+  MinLevelThreshold:    256
+  SamplesType:          UnKnownCodec
+
+Base VADClipping:
+  SamplesType:          UnKnownCodec
+
+Base Noise:
+  Interval:             0.1    # Seconds
+  DetectorType:         RMS    # This can be FFT as well
+  NoiseThreshold:       20
+  SignalThreshold:      80
+  Normalize:            no
+  RemoveBias:           no
+  ResultDb:             yes
+  WindowType:           Hann
+  WindowWidth:          3
+
+# Moved to Override: sections
+# DeadAir-01:
+#    MinLevelThreshold:    0
+
+# SilentCall:
+#    MinLevelThreshold:    0
+#    IsUseRMSPower:        yes
+#    MinRMSThreshold:      -70
+
+# Dynamic-Clipping:
+#    FlyAddingCoefficient: 0.1000
+#    SamplesType:          UnKnownCodec
+#    IsUseDynamicClipping: yes
+
+Correction:
+  - IntStart:     5.0
+    IntEnd:       4.2
+    Mult:         1.0
+    Shift:        0
+
+  - IntStart:     4.2
+    IntEnd:       3.5
+    Mult:         1.0
+    Shift:        0
+
+
+SR Correction:
+  - SampleRate:     11000.0
+    Shift:          0.05
+
+  - SampleRate:     16000.0
+    Shift:          0.1
+
+  - SampleRate:     22000.0
+    Shift:          0.2
+
+  - SampleRate:     32000.0
+    Shift:          0.3
+
+  - SampleRate:     48000.0
+    Shift:          0.45
+
+  - SampleRate:     96000.0
+    Shift:          0.5
+
+  - SampleRate:     192000.0
+    Shift:          0.6
+
+Scores Map:
+  ScoresLine:       4;3.027000;2.935000;2.905000;2.818000;2.590000;2.432000;2.310000;1.665000;1.000000;
 

src/agent_gsm.py

@@ -152,17 +152,20 @@ def run_analyze(file_test: str, file_reference: str, number: str) -> bool:
             return False
         
         try:
-            bounds_signal : SignalBoundaries = detect_degraded_signal(Path(file_test), Path(file_reference))
-            # bounds_signal.offset_start = 0
-            # bounds_signal.offset_finish = 0
-            print(f'Found signal bounds: {bounds_signal}')
-
+            bounds_signal : SignalBoundaries = None
+            if is_caller:
+                bounds_signal.offset_start = 10.0   # Skip ringtones
+                bounds_signal.offset_finish = 1.0   # Eat possible end tone
+            elif is_answerer:
+                bounds_signal.offset_start = 0.0
+                bounds_signal.offset_finish = 1.0   # Eat possible end tone
+                
             # PVQA report
             pvqa_mos, pvqa_report, pvqa_rfactor = utils_sevana.find_pvqa_mos(file_test, bounds_signal.offset_start, bounds_signal.offset_finish)
             utils.log(f'PVQA MOS: {pvqa_mos}, PVQA R-factor: {pvqa_rfactor}')
 
             # AQuA report            
-            bounds_reference : SignalBoundaries = detect_reference_signal(Path(file_reference))
+            bounds_reference : SignalBoundaries = SignalBoundaries()
             bounds_reference.offset_start = 0
             bounds_reference.offset_finish = 0
 
@@ -258,7 +261,7 @@ def make_call(target: str):
                                timelimit_seconds=ref_time_length, 
                                target=target)
         
-        run_analyze(CONFIG.RecordFile, CONFIG.ReferenceAudio, target)
+        run_analyze(CONFIG.RecordFile, CONFIG.PreparedReferenceAudio, target)
     except Exception as e:
         utils.log_error(f'BT I/O failed finally. Error: {str(e)}')
                         

src/utils_sevana.py

@@ -12,8 +12,6 @@ import time
 import urllib
 
 from pathlib import Path
-from colorama import Fore, Style
-from utils_cache import InfoCache
 
 PVQA_CMD = "{pvqa} --license {pvqa_lic} --config {pvqa_cfg} --mode analysis --channel 0 " \
            "--report {output} --input {input} --cut-begin {cut_begin} --cut-end {cut_end}"
@@ -21,10 +19,8 @@ PVQA_CMD = "{pvqa} --license {pvqa_lic} --config {pvqa_cfg} --mode analysis --ch
 PVQA_CMD_LIC_SERVER = "{pvqa} --license-server {pvqa_lic} --config {pvqa_cfg} --mode analysis --channel 0 " \
                       "--report {output} --input {input}"
 
-AQUA_CMD = ("{aqua} {aqua_lic} -mode files -src file \"{reference}\" -tstf \"{input}\" -avlp off -smtnrm on "
-            "-decor off -mprio off -acr auto -npnt auto -voip on -enorm rms -g711 off "
-            "-spfrcor on -grad off -tmc on -hist-pitch on on -hist-levels on on on -miter 1 -specp 32 {spectrum}  "
-            "-ratem %%m -fau {faults} -output json -trim r 15 -cut-tst {cut_begin} {cut_end} -cut-src {cut_begin_src} {cut_end_src}")
+AQUA_CMD = ("{aqua} {aqua_lic} -mode files -src file \"{reference}\" -tstf \"{input}\" -config {aqua_config} "
+            "-specp 32 {spectrum} -fau {faults} -cut-tst {cut_begin} {cut_end} -cut-src {cut_begin_src} {cut_end_src}")
 
 PVQA_PATH = ""
 PVQA_LIC_PATH = "pvqa.lic"
@@ -32,6 +28,7 @@ PVQA_CFG_PATH = "pvqa.cfg"
 
 AQUA_PATH = ""
 AQUA_LIC_PATH = "aqua-wb.lic"
+AQUA_CFG_PATH = "aqua.cfg"
 
 SILER_PATH = ""
 
@@ -78,9 +75,11 @@ def load_config_and_licenses(server: str):
         load_file(utils.join_host_and_path(server, '/deploy/pvqa.cfg'), PVQA_CFG_PATH)
         load_file(utils.join_host_and_path(server, '/deploy/pvqa.lic'), PVQA_LIC_PATH)
         load_file(utils.join_host_and_path(server, '/deploy/aqua-wb.lic'), AQUA_LIC_PATH)
+        load_file(utils.join_host_and_path(server, '/deploy/aqua.cfg'), AQUA_CFG_PATH)
     except Exception as e:
         utils.log_error(f'Failed to fetch new licenses and config. Skipping it.')
 
+
 def find_binaries(bin_directory: Path, license_server: str = None) -> bool:
     # Update path to pvqa/aqua-wb
     global PVQA_CFG_PATH, PVQA_LIC_PATH, AQUA_LIC_PATH, PVQA_PATH, AQUA_PATH, PVQA_CMD, AQUA_CMD, SILER_PATH, SPEECH_DETECTOR_PATH
@@ -94,6 +93,7 @@ def find_binaries(bin_directory: Path, license_server: str = None) -> bool:
     PVQA_LIC_PATH =         bin_directory / PVQA_LIC_PATH
     PVQA_CFG_PATH =         bin_directory / PVQA_CFG_PATH
     AQUA_PATH =             bin_directory / platform_prefix / AQUA_PATH
+    AQUA_CFG_PATH =         bin_directory / AQUA_CFG_PATH
     AQUA_LIC_PATH =         bin_directory / AQUA_LIC_PATH
     SILER_PATH =            bin_directory / platform_prefix / SILER_PATH
     SPEECH_DETECTOR_PATH =  bin_directory / platform_prefix / SPEECH_DETECTOR_PATH
@@ -109,9 +109,16 @@ def find_binaries(bin_directory: Path, license_server: str = None) -> bool:
         PVQA_CFG_PATH = Path(utils.get_script_path()) / 'pvqa.cfg'
     
     if not PVQA_CFG_PATH.exists():
-        utils.log_error(f'Failed to find pvqa config.')
+        utils.log_error(f'Failed to find PVQA config file.')
         return False
 
+    if not AQUA_CFG_PATH.exists():
+        AQUA_CFG_PATH = Path(utils.get_script_path()) / 'aqua.cfg'
+
+    if not AQUA_CFG_PATH.exists():
+        utils.log_error(f'Failed to find AQuA config file.')
+        return False
+    
     if not AQUA_PATH.exists():
         utils.log_error(f'Failed to find aqua-wb binary.')
         return False
@@ -238,7 +245,7 @@ def find_aqua_mos(good_path, test_path, test_file_offset_begin: float = 0.0, tes
                   good_file_offset_begin: float = 0.0, good_file_offset_end: float = 0.0):
     try:
         out_data = ""
-        cmd = AQUA_CMD.format(aqua=AQUA_PATH, aqua_lic=AQUA_LIC_PATH,
+        cmd = AQUA_CMD.format(aqua=AQUA_PATH, aqua_lic=AQUA_LIC_PATH, aqua_config = AQUA_CFG_PATH,
                               reference=good_path, input=test_path, spectrum=AQUA_SPECTRUM,
                               faults=AQUA_FAULTS, 
                               cut_begin=int(test_file_offset_begin * 1000), cut_end=int(test_file_offset_end * 1000),