fix spectrum_type typo

deepspeech/exps/deepspeech2/config.py

@@ -32,7 +32,7 @@ _C.data = CN(
         window_ms=20.0,  # ms
         n_fft=None,  # fft points
         max_freq=None,  # None for samplerate/2
-        specgram_type='linear',  # 'linear', 'mfcc', 'fbank'
+        spectrum_type='linear',  # 'linear', 'mfcc', 'fbank'
         feat_dim=0,  # 'mfcc', 'fbank'
         delat_delta=False,  # 'mfcc', 'fbank'
         target_sample_rate=16000,  # target sample rate

deepspeech/frontend/featurizer/audio_featurizer.py

@@ -24,15 +24,15 @@ class AudioFeaturizer(object):
 
     Currently, it supports feature types of linear spectrogram and mfcc.
 
-    :param specgram_type: Specgram feature type. Options: 'linear'.
-    :type specgram_type: str
+    :param spectrum_type: Specgram feature type. Options: 'linear'.
+    :type spectrum_type: str
     :param stride_ms: Striding size (in milliseconds) for generating frames.
     :type stride_ms: float
     :param window_ms: Window size (in milliseconds) for generating frames.
     :type window_ms: float
-    :param max_freq: When specgram_type is 'linear', only FFT bins
+    :param max_freq: When spectrum_type is 'linear', only FFT bins
                      corresponding to frequencies between [0, max_freq] are
-                     returned; when specgram_type is 'mfcc', max_feq is the
+                     returned; when spectrum_type is 'mfcc', max_feq is the
                      highest band edge of mel filters.
     :types max_freq: None|float
     :param target_sample_rate: Audio are resampled (if upsampling or
@@ -47,7 +47,7 @@ class AudioFeaturizer(object):
     """
 
     def __init__(self,
-                 specgram_type: str='linear',
+                 spectrum_type: str='linear',
                  feat_dim: int=None,
                  delta_delta: bool=False,
                  stride_ms=10.0,
@@ -58,7 +58,7 @@ class AudioFeaturizer(object):
                  use_dB_normalization=True,
                  target_dB=-20,
                  dither=1.0):
-        self._specgram_type = specgram_type
+        self._spectrum_type = spectrum_type
         # mfcc and fbank using `feat_dim`
         self._feat_dim = feat_dim
         # mfcc and fbank using `delta-delta`
@@ -113,27 +113,27 @@ class AudioFeaturizer(object):
     def feature_size(self):
         """audio feature size"""
         feat_dim = 0
-        if self._specgram_type == 'linear':
+        if self._spectrum_type == 'linear':
             fft_point = self._window_ms if self._fft_point is None else self._fft_point
             feat_dim = int(fft_point * (self._target_sample_rate / 1000) / 2 +
                            1)
-        elif self._specgram_type == 'mfcc':
+        elif self._spectrum_type == 'mfcc':
             # mfcc, delta, delta-delta
             feat_dim = int(self._feat_dim *
                            3) if self._delta_delta else int(self._feat_dim)
-        elif self._specgram_type == 'fbank':
+        elif self._spectrum_type == 'fbank':
             # fbank, delta, delta-delta
             feat_dim = int(self._feat_dim *
                            3) if self._delta_delta else int(self._feat_dim)
         else:
-            raise ValueError("Unknown specgram_type %s. "
-                             "Supported values: linear." % self._specgram_type)
+            raise ValueError("Unknown spectrum_type %s. "
+                             "Supported values: linear." % self._spectrum_type)
         return feat_dim
 
     def _compute_specgram(self, audio_segment):
         """Extract various audio features."""
         sample_rate = audio_segment.sample_rate
-        if self._specgram_type == 'linear':
+        if self._spectrum_type == 'linear':
             samples = audio_segment.samples
             return self._compute_linear_specgram(
                 samples,
@@ -141,7 +141,7 @@ class AudioFeaturizer(object):
                 stride_ms=self._stride_ms,
                 window_ms=self._window_ms,
                 max_freq=self._max_freq)
-        elif self._specgram_type == 'mfcc':
+        elif self._spectrum_type == 'mfcc':
             samples = audio_segment.to('int16')
             return self._compute_mfcc(
                 samples,
@@ -152,7 +152,7 @@ class AudioFeaturizer(object):
                 max_freq=self._max_freq,
                 dither=self._dither,
                 delta_delta=self._delta_delta)
-        elif self._specgram_type == 'fbank':
+        elif self._spectrum_type == 'fbank':
             samples = audio_segment.to('int16')
             return self._compute_fbank(
                 samples,
@@ -164,8 +164,8 @@ class AudioFeaturizer(object):
                 dither=self._dither,
                 delta_delta=self._delta_delta)
         else:
-            raise ValueError("Unknown specgram_type %s. "
-                             "Supported values: linear." % self._specgram_type)
+            raise ValueError("Unknown spectrum_type %s. "
+                             "Supported values: linear." % self._spectrum_type)
 
     def _compute_linear_specgram(self,
                                  samples,

deepspeech/frontend/featurizer/speech_featurizer.py

@@ -27,16 +27,16 @@ class SpeechFeaturizer(object):
 
     :param vocab_filepath: Filepath to load vocabulary for token indices
                            conversion.
-    :type specgram_type: str
-    :param specgram_type: Specgram feature type. Options: 'linear', 'mfcc'.
-    :type specgram_type: str
+    :type spectrum_type: str
+    :param spectrum_type: Specgram feature type. Options: 'linear', 'mfcc'.
+    :type spectrum_type: str
     :param stride_ms: Striding size (in milliseconds) for generating frames.
     :type stride_ms: float
     :param window_ms: Window size (in milliseconds) for generating frames.
     :type window_ms: float
-    :param max_freq: When specgram_type is 'linear', only FFT bins
+    :param max_freq: When spectrum_type is 'linear', only FFT bins
                      corresponding to frequencies between [0, max_freq] are
-                     returned; when specgram_type is 'mfcc', max_freq is the
+                     returned; when spectrum_type is 'mfcc', max_freq is the
                      highest band edge of mel filters.
     :types max_freq: None|float
     :param target_sample_rate: Speech are resampled (if upsampling or
@@ -54,7 +54,7 @@ class SpeechFeaturizer(object):
                  unit_type,
                  vocab_filepath,
                  spm_model_prefix=None,
-                 specgram_type='linear',
+                 spectrum_type='linear',
                  feat_dim=None,
                  delta_delta=False,
                  stride_ms=10.0,
@@ -66,7 +66,7 @@ class SpeechFeaturizer(object):
                  target_dB=-20,
                  dither=1.0):
         self._audio_featurizer = AudioFeaturizer(
-            specgram_type=specgram_type,
+            spectrum_type=spectrum_type,
             feat_dim=feat_dim,
             delta_delta=delta_delta,
             stride_ms=stride_ms,

deepspeech/io/__init__.py

@@ -35,7 +35,7 @@ def create_dataloader(manifest_path,
                       stride_ms=10.0,
                       window_ms=20.0,
                       max_freq=None,
-                      specgram_type='linear',
+                      spectrum_type='linear',
                       feat_dim=None,
                       delta_delta=False,
                       use_dB_normalization=True,
@@ -64,7 +64,7 @@ def create_dataloader(manifest_path,
         stride_ms=stride_ms,
         window_ms=window_ms,
         max_freq=max_freq,
-        specgram_type=specgram_type,
+        spectrum_type=spectrum_type,
         feat_dim=feat_dim,
         delta_delta=delta_delta,
         use_dB_normalization=use_dB_normalization,

deepspeech/io/dataset.py

@@ -63,7 +63,7 @@ class ManifestDataset(Dataset):
                 n_fft=None,  # fft points
                 max_freq=None,  # None for samplerate/2
                 raw_wav=True,  # use raw_wav or kaldi feature
-                specgram_type='linear',  # 'linear', 'mfcc', 'fbank'
+                spectrum_type='linear',  # 'linear', 'mfcc', 'fbank'
                 feat_dim=0,  # 'mfcc', 'fbank'
                 delta_delta=False,  # 'mfcc', 'fbank'
                 dither=1.0,  # feature dither
@@ -124,7 +124,7 @@ class ManifestDataset(Dataset):
             n_fft=config.data.n_fft,
             max_freq=config.data.max_freq,
             target_sample_rate=config.data.target_sample_rate,
-            specgram_type=config.data.specgram_type,
+            spectrum_type=config.data.spectrum_type,
             feat_dim=config.data.feat_dim,
             delta_delta=config.data.delta_delta,
             dither=config.data.dither,
@@ -152,7 +152,7 @@ class ManifestDataset(Dataset):
                  n_fft=None,
                  max_freq=None,
                  target_sample_rate=16000,
-                 specgram_type='linear',
+                 spectrum_type='linear',
                  feat_dim=None,
                  delta_delta=False,
                  dither=1.0,
@@ -180,7 +180,7 @@ class ManifestDataset(Dataset):
             n_fft (int, optional): fft points for rfft. Defaults to None.
             max_freq (int, optional): max cut freq. Defaults to None.
             target_sample_rate (int, optional): target sample rate which used for training. Defaults to 16000.
-            specgram_type (str, optional): 'linear', 'mfcc' or 'fbank'. Defaults to 'linear'.
+            spectrum_type (str, optional): 'linear', 'mfcc' or 'fbank'. Defaults to 'linear'.
             feat_dim (int, optional): audio feature dim, using by 'mfcc' or 'fbank'. Defaults to None.
             delta_delta (bool, optional): audio feature with delta-delta, using by 'fbank' or 'mfcc'. Defaults to False.
             use_dB_normalization (bool, optional): do dB normalization. Defaults to True.
@@ -200,7 +200,7 @@ class ManifestDataset(Dataset):
             unit_type=unit_type,
             vocab_filepath=vocab_filepath,
             spm_model_prefix=spm_model_prefix,
-            specgram_type=specgram_type,
+            spectrum_type=spectrum_type,
             feat_dim=feat_dim,
             delta_delta=delta_delta,
             stride_ms=stride_ms,

examples/aishell/s0/conf/deepspeech2.yaml

@@ -13,7 +13,7 @@ data:
   max_output_len: .inf
   min_output_input_ratio: 0.00
   max_output_input_ratio: .inf
-  specgram_type: linear
+  spectrum_type: linear
   target_sample_rate: 16000
   max_freq: None
   n_fft: None

examples/aishell/s0/local/data.sh

@@ -46,7 +46,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
     num_workers=$(nproc)
     python3 ${MAIN_ROOT}/utils/compute_mean_std.py \
     --manifest_path="data/manifest.train.raw" \
-    --specgram_type="linear" \
+    --spectrum_type="linear" \
     --delta_delta=false \
     --stride_ms=10.0 \
     --window_ms=20.0 \

examples/aishell/s1/conf/chunk_conformer.yaml

@@ -15,7 +15,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/aishell/s1/conf/conformer.yaml

@@ -15,7 +15,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/aishell/s1/local/data.sh

@@ -46,7 +46,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
     num_workers=$(nproc)
     python3 ${MAIN_ROOT}/utils/compute_mean_std.py \
     --manifest_path="data/manifest.train.raw" \
-    --specgram_type="fbank" \
+    --spectrum_type="fbank" \
     --feat_dim=80 \
     --delta_delta=false \
     --stride_ms=10.0 \

examples/librispeech/s0/conf/deepspeech2.yaml

@@ -13,7 +13,7 @@ data:
   max_output_len: .inf
   min_output_input_ratio: 0.00
   max_output_input_ratio: .inf
-  specgram_type: linear
+  spectrum_type: linear
   target_sample_rate: 16000
   max_freq: None
   n_fft: None

examples/librispeech/s0/local/data.sh

@@ -62,7 +62,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
     python3 ${MAIN_ROOT}/utils/compute_mean_std.py \
     --manifest_path="data/manifest.train.raw" \
     --num_samples=2000 \
-    --specgram_type="linear" \
+    --spectrum_type="linear" \
     --delta_delta=false \
     --sample_rate=16000 \
     --stride_ms=10.0 \

examples/librispeech/s1/conf/chunk_confermer.yaml

@@ -16,7 +16,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/librispeech/s1/conf/chunk_transformer.yaml

@@ -16,7 +16,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/librispeech/s1/conf/conformer.yaml

@@ -16,7 +16,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/librispeech/s1/conf/transformer.yaml

@@ -16,7 +16,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/librispeech/s1/local/data.sh

@@ -68,7 +68,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
     python3 ${MAIN_ROOT}/utils/compute_mean_std.py \
     --manifest_path="data/manifest.train.raw" \
     --num_samples=-1 \
-    --specgram_type="fbank" \
+    --spectrum_type="fbank" \
     --feat_dim=80 \
     --delta_delta=false \
     --sample_rate=16000 \

examples/tiny/s0/conf/deepspeech2.yaml

@@ -13,7 +13,7 @@ data:
   max_output_len: 400.0
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
-  specgram_type: linear
+  spectrum_type: linear
   target_sample_rate: 16000
   max_freq: None
   n_fft: None

examples/tiny/s0/local/data.sh

-#! /usr/bin/env bash
+#!/bin/bash
 
 stage=-1
 stop_stage=100
@@ -46,7 +46,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
     python3 ${MAIN_ROOT}/utils/compute_mean_std.py \
     --manifest_path="data/manifest.tiny.raw" \
     --num_samples=64 \
-    --specgram_type="linear" \
+    --spectrum_type="linear" \
     --delta_delta=false \
     --sample_rate=16000 \
     --stride_ms=10.0 \

examples/tiny/s1/conf/chunk_confermer.yaml

@@ -16,7 +16,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/tiny/s1/conf/chunk_transformer.yaml

@@ -16,7 +16,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/tiny/s1/conf/conformer.yaml

@@ -16,7 +16,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/tiny/s1/conf/transformer.yaml

@@ -16,7 +16,7 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  spectrum_type: fbank #linear, mfcc, fbank
   feat_dim: 80
   delta_delta: False
   dither: 1.0

examples/tiny/s1/local/data.sh

@@ -51,7 +51,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
     python3 ${MAIN_ROOT}/utils/compute_mean_std.py \
     --manifest_path="data/manifest.tiny.raw" \
     --num_samples=64 \
-    --specgram_type="fbank" \
+    --spectrum_type="fbank" \
     --feat_dim=80 \
     --delta_delta=false \
     --sample_rate=16000 \