def prepare_processing_graph(self, model_settings):
"""Builds a TensorFlow graph to apply the input distortions."""
desired_samples = model_settings['desired_samples']
self.wav_filename_placeholder_ = tf.placeholder(tf.string, [],
name='filename')
wav_loader = tf.io.read_file(self.wav_filename_placeholder_)
wav_decoder = tf.audio.decode_wav(wav_loader,
desired_channels=1,
desired_samples=desired_samples)
# Allow the audio sample's volume to be adjusted.
self.foreground_volume_placeholder_ = tf.placeholder(
tf.float32, [], name='foreground_volme')
scaled_foreground = tf.multiply(wav_decoder.audio,
self.foreground_volume_placeholder_)
# Shift the sample's start position, and pad any gaps with zeros.
self.time_shift_placeholder_ = tf.placeholder(tf.int32,
name='timeshift')
shifted_foreground = tf_roll(scaled_foreground,
self.time_shift_placeholder_)
# Mix in background noise.
self.background_data_placeholder_ = tf.placeholder(
tf.float32, [desired_samples, 1], name='background_data')
self.background_volume_placeholder_ = tf.placeholder(
tf.float32, [], name='background_volume')
background_mul = tf.multiply(self.background_data_placeholder_,
self.background_volume_placeholder_)
background_add = tf.add(background_mul, shifted_foreground)
# removed clipping: tf.clip_by_value(background_add, -1.0, 1.0)
self.background_clamp_ = background_add
self.background_clamp_ = tf.reshape(
self.background_clamp_, (1, model_settings['desired_samples']))
# Run the spectrogram and MFCC ops to get a 2D 'fingerprint' of the audio.
stfts = tf.signal.stft(
self.background_clamp_,
frame_length=model_settings['window_size_samples'],
frame_step=model_settings['window_stride_samples'],
fft_length=None)
self.spectrogram_ = tf.abs(stfts)
print("self", self.spectrogram_.shape)
num_spectrogram_bins = self.spectrogram_.shape[-1]
lower_edge_hertz, upper_edge_hertz = 80.0, 7600.0
linear_to_mel_weight_matrix = \
tf.signal.linear_to_mel_weight_matrix(
model_settings['dct_coefficient_count'],
num_spectrogram_bins, model_settings['sample_rate'],
lower_edge_hertz, upper_edge_hertz)
mel_spectrograms = tf.tensordot(self.spectrogram_,
linear_to_mel_weight_matrix, 1)
mel_spectrograms.set_shape(self.spectrogram_.shape[:-1].concatenate(
linear_to_mel_weight_matrix.shape[-1:]))
log_mel_spectrograms = tf.log(mel_spectrograms + 1e-6)
self.mfcc_ = tf.signal.mfccs_from_log_mel_spectrograms(
log_mel_spectrograms
)[:, :, :model_settings['num_log_mel_features']] # :13
def prepare_processing_graph(self, model_settings):
"""Builds a TensorFlow graph to apply the input distortions"""
desired_samples = model_settings['desired_samples']
self.wav_filename_placeholder_ = tf.placeholder(
tf.string, [], name='filename')
wav_loader = io_ops.read_file(self.wav_filename_placeholder_)
wav_decoder = contrib_audio.decode_wav(
wav_loader, desired_channels=1, desired_samples=desired_samples)
# Allow the audio sample's volume to be adjusted.
self.foreground_volume_placeholder_ = tf.placeholder(
tf.float32, [], name='foreground_volme')
scaled_foreground = tf.multiply(wav_decoder.audio,
self.foreground_volume_placeholder_)
# Shift the sample's start position, and pad any gaps with zeros.
self.time_shift_placeholder_ = tf.placeholder(tf.int32, name='timeshift')
shifted_foreground = tf_roll(scaled_foreground,
self.time_shift_placeholder_)
# Mix in background noise.
self.background_data_placeholder_ = tf.placeholder(
tf.float32, [desired_samples, 1], name='background_data')
self.background_volume_placeholder_ = tf.placeholder(
tf.float32, [], name='background_volume')
background_mul = tf.multiply(self.background_data_placeholder_,
self.background_volume_placeholder_)
background_add = tf.add(background_mul, shifted_foreground)
# removed clipping: tf.clip_by_value(background_add, -1.0, 1.0)
self.background_clamp_ = background_add
self.background_clamp_ = tf.reshape(self.background_clamp_,
(1, model_settings['desired_samples']))
# Run the spectrogram and MFCC ops to get a 2D 'fingerprint' of the audio.
stfts = tf.contrib.signal.stft(
self.background_clamp_,
frame_length=model_settings['window_size_samples'],
frame_step=model_settings['window_stride_samples'],
fft_length=None)
self.spectrogram_ = tf.abs(stfts)
num_spectrogram_bins = self.spectrogram_.shape[-1].value
lower_edge_hertz, upper_edge_hertz = 80.0, 7600.0
linear_to_mel_weight_matrix = \
tf.contrib.signal.linear_to_mel_weight_matrix(
model_settings['dct_coefficient_count'],
num_spectrogram_bins, model_settings['sample_rate'],
lower_edge_hertz, upper_edge_hertz)
mel_spectrograms = tf.tensordot(self.spectrogram_,
linear_to_mel_weight_matrix, 1)
mel_spectrograms.set_shape(self.spectrogram_.shape[:-1].concatenate(
linear_to_mel_weight_matrix.shape[-1:]))
log_mel_spectrograms = tf.log(mel_spectrograms + 1e-6)
self.mfcc_ = tf.contrib.signal.mfccs_from_log_mel_spectrograms(
log_mel_spectrograms)[:, :, :
model_settings['num_log_mel_features']] # :13
def prepare_processing_graph(self, model_settings):
"""Builds a TensorFlow graph to apply the input distortions.
Creates a graph that loads a WAVE file, decodes it, scales the volume,
shifts it in time, adds in background noise, calculates a spectrogram, and
then builds an MFCC fingerprint from that.
This must be called with an active TensorFlow session running, and it
creates multiple placeholder inputs, and one output:
- wav_filename_placeholder_: Filename of the WAV to load.
- foreground_volume_placeholder_: How loud the main clip should be.
- time_shift_placeholder_: How much the clip is shifted.
- background_data_placeholder_: PCM sample data for background noise.
- background_volume_placeholder_: Loudness of mixed-in background.
- mfcc_: Output 2D fingerprint of processed audio.
Args:
model_settings: Information about the current model being trained.
"""
desired_samples = model_settings['desired_samples']
self.wav_filename_placeholder_ = tf.placeholder(tf.string, [],
name='filename')
wav_loader = io_ops.read_file(self.wav_filename_placeholder_)
wav_decoder = contrib_audio.decode_wav(wav_loader,
desired_channels=1,
desired_samples=desired_samples)
# Allow the audio sample's volume to be adjusted.
self.foreground_volume_placeholder_ = tf.placeholder(
tf.float32, [], name='foreground_volme')
scaled_foreground = tf.multiply(wav_decoder.audio,
self.foreground_volume_placeholder_)
# Shift the sample's start position, and pad any gaps with zeros.
self.time_shift_placeholder_ = tf.placeholder(tf.int32,
name='timeshift')
# TODO(see--): Write test with np.roll
shifted_foreground = tf_roll(scaled_foreground,
self.time_shift_placeholder_)
# Mix in background noise.
self.background_data_placeholder_ = tf.placeholder(
tf.float32, [desired_samples, 1], name='background_data')
self.background_volume_placeholder_ = tf.placeholder(
tf.float32, [], name='background_volume')
background_mul = tf.multiply(self.background_data_placeholder_,
self.background_volume_placeholder_)
background_add = tf.add(background_mul, shifted_foreground)
# removed clipping: tf.clip_by_value(background_add, -1.0, 1.0)
self.background_clamp_ = background_add
self.background_clamp_ = tf.reshape(
self.background_clamp_, (1, model_settings['desired_samples']))
# Run the spectrogram and MFCC ops to get a 2D 'fingerprint' of the audio.
stfts = tf.contrib.signal.stft(
self.background_clamp_,
frame_length=model_settings['window_size_samples'],
frame_step=model_settings['window_stride_samples'],
fft_length=None)
self.spectrogram_ = tf.abs(stfts)
num_spectrogram_bins = self.spectrogram_.shape[-1].value
lower_edge_hertz, upper_edge_hertz = 80.0, 7600.0
linear_to_mel_weight_matrix = \
tf.contrib.signal.linear_to_mel_weight_matrix(
model_settings['dct_coefficient_count'],
num_spectrogram_bins, model_settings['sample_rate'],
lower_edge_hertz, upper_edge_hertz)
mel_spectrograms = tf.tensordot(self.spectrogram_,
linear_to_mel_weight_matrix, 1)
mel_spectrograms.set_shape(self.spectrogram_.shape[:-1].concatenate(
linear_to_mel_weight_matrix.shape[-1:]))
log_mel_spectrograms = tf.log(mel_spectrograms + 1e-6)
self.mfcc_ = tf.contrib.signal.mfccs_from_log_mel_spectrograms(
log_mel_spectrograms
)[:, :, :model_settings['num_log_mel_features']] # :13