def testConstantInput(self):
beatPeriod = 21168 # N.B The beat period is 21168 samples for 125 bpm @ 44.1k samp. rate
onesAudio = ones(beatPeriod)
bpmEstimate = 0
confidence = LoopBpmConfidence()(onesAudio, bpmEstimate)
self.assertEquals(0, confidence)
# Non-zero estimate check results in non-zero confidence
# The estimation is based on audio length.
# Different constant input length will result in different estimations.
bpmEstimate = 125
confidence = LoopBpmConfidence()(onesAudio, bpmEstimate)
self.assertEquals(1, confidence)
def testEmpty(self):
# Zero estimate check results in zero confidence
emptyAudio = []
bpmEstimate = 0
confidence = LoopBpmConfidence()(emptyAudio, bpmEstimate)
self.assertEquals(0, confidence)
# Non-zero estimate check results in zero confidence
# The estimation is based on audio length.
# Different constant input length will result in different estimations.
emptyAudio = []
bpmEstimate = 125
confidence = LoopBpmConfidence()(emptyAudio, bpmEstimate)
self.assertEquals(0, confidence)
def testZero(self):
beatPeriod = 21168 # N.B The beat period is 21168 samples for 125 bpm @ 44.1k samp. rate
zeroAudio = zeros(beatPeriod)
bpmEstimate = 0
confidence = LoopBpmConfidence()(zeroAudio, bpmEstimate)
self.assertEquals(0, confidence)
# Non-zero estimate check results in non-zero confidence
# The estimation is based on audio length.
# Different constant input length will result in different estimations.
bpmEstimate = 125
confidence = LoopBpmConfidence()(zeroAudio, bpmEstimate)
self.assertNotEquals(0, confidence)
# A silent length of 4 Beat periods produces a confidence of 1.
zeroAudio4Beats = zeros(beatPeriod*4)
bpmEstimate = 125
confidence = LoopBpmConfidence()(zeroAudio4Beats, bpmEstimate)
self.assertEquals(1, confidence)
def testExactAudioLengthMatch(self):
audio = MonoLoader(filename=join(testdata.audio_dir, 'recorded', 'techno_loop.wav'))()
bpmEstimate = 125
beatPeriod = 21168 # N.B The beat period is 21168 samples for 125 bpm @ 44.1k samp. rate
silentAudio = zeros(beatPeriod)
# Add non-musical silence to the beginning of the audio
signal1 = numpy.append(silentAudio, audio)
confidence = LoopBpmConfidence()(signal1, bpmEstimate)
self.assertEquals(confidence, 1.0)
# Add non-musical silence to the end of the audio
signal2 = numpy.append(audio, silentAudio)
confidence = LoopBpmConfidence()(signal2, bpmEstimate)
self.assertEquals(confidence, 1.0)
# Concatenate silence at both ends
signal3 = numpy.append(signal1, silentAudio)
confidence = LoopBpmConfidence()(signal3, bpmEstimate)
self.assertEquals(confidence, 1.0)
def testSilentEdge(self):
audio = MonoLoader(filename=join(testdata.audio_dir, 'recorded', 'techno_loop.wav'))()
bpmEstimate = 125
lenSilence = 30000 # N.B The beat period is 21168 samples for 125 bpm @ 44.1k samp. rate
silentAudio = zeros(lenSilence)
benchmarkConfidence = 0.96 # This figure was arrived at emperically from the min. confidence observed with test runs
# Test addition of non-musical silence before the loop starts
# The length is not a beat period,
# Nonetheless, we can stillreliably estimate the starting point because it is a hard transient.
signal1 = numpy.append(silentAudio, audio)
confidence = LoopBpmConfidence()(signal1, bpmEstimate)
self.assertGreater(confidence, benchmarkConfidence)
def testRegression(self):
audio = MonoLoader(filename=join(testdata.audio_dir, 'recorded', 'techno_loop.wav'))()
"""
Confidence figures are obtained from previous runs algorithm as of
code baseline on 10-Feb-2021
In keeping with regression tests principles, no future changes of the algorithm
should break these tests.
Test for different BPMs starting with correct one: 125,
and check the confidence levels obtained from previous runs.
The 3rd param in assertAlmostEqual() has been optimised to 0.01.
We chose this reasonable margin since the expected confidence benchmark
has been rounded to two places of decimal.
"""
bpmEstimate = 125
expectedConfidence = 1
confidence = LoopBpmConfidence()(audio, bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
bpmEstimate = 120
expectedConfidence = 0.87
confidence = LoopBpmConfidence()(audio, bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
bpmEstimate = 70
expectedConfidence = 0.68
confidence = LoopBpmConfidence()(audio, bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
"""
Apply same principle above but now to Sub-Sections of Audio,
truncating a bit at the end
The following confidence measurements were taken on 11 Feb. 2021
for a given set of parameters on different audio subsections.
A segment of techno_loop gives a non-ideal estimation of BPM
so we expect the confidence to drop.
Subsection [0:len90]
BPM Estimate: 125: Confidence: 0.19992440938949585
BPM Estimate: 120: Confidence: 0.4305669069290161
BPM Estimate: 70: Confidence: 0.5011640191078186
Subsection [len01:len90],
BPM Estimate: 125: Confidence: 0.9199735522270203
BPM Estimate: 120: Confidence: 0.3631746172904968
BPM Estimate: 70: Confidence: 0.7951852083206177
The ASSERTs will check for these to two places of decimal
These checks act as integrity checks. If any future changes break these asserts,
then this will indicate that accuracy of algorithm has changed.
The 3rd param in assertAlmostEqual() has been optimised, lowest possible value chosen.
"""
# Define Markers for significant meaningful subsections
# consistent with sampling rate
len90 = int(0.9*len(audio)) # End point for 90% loop and subsection
len01 = int(0.01*len(audio)) # Begin point for subsection loop
# Truncated Audios 90%
bpmEstimate = 125
expectedConfidence = 0.2 # rounded from measured 0.19992440938949585
confidence = LoopBpmConfidence()(audio[0:len90], bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
bpmEstimate = 120
expectedConfidence = 0.43 # rounded from measured 0.4305669069290161
confidence = LoopBpmConfidence()(audio[0:len90], bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
bpmEstimate = 70
expectedConfidence = 0.50 # rounded from measured 0.5011640191078186
confidence = LoopBpmConfidence()(audio[0:len90], bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
# Truncated Audios 90% and 1% at beginning
bpmEstimate = 125
expectedConfidence = 0.92 # rounded from measured 0.9199735522270203
confidence = LoopBpmConfidence()(audio[len01:len90], bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
bpmEstimate = 120
expectedConfidence = 0.36 # rounded from measured 0.3631746172904968
confidence = LoopBpmConfidence()(audio[len01:len90], bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
bpmEstimate = 70
expectedConfidence = 0.79 # rounded from measured 0.7951852083206177
confidence = LoopBpmConfidence()(audio[len01:len90], bpmEstimate)
self.assertAlmostEqual(expectedConfidence, confidence, 0.01)
def testInvalidParam(self):
self.assertConfigureFails(LoopBpmConfidence(), {'sampleRate': -1})