diff options
-rw-r--r-- | tensorflow/core/ops/compat/ops_history.v1.pbtxt | 43 | ||||
-rw-r--r-- | tensorflow/core/ops/ops.pbtxt | 51 |
2 files changed, 94 insertions, 0 deletions
diff --git a/tensorflow/core/ops/compat/ops_history.v1.pbtxt b/tensorflow/core/ops/compat/ops_history.v1.pbtxt index 50e4971314..1781f778b4 100644 --- a/tensorflow/core/ops/compat/ops_history.v1.pbtxt +++ b/tensorflow/core/ops/compat/ops_history.v1.pbtxt @@ -11317,6 +11317,49 @@ op { } } op { + name: "Mfcc" + input_arg { + name: "spectrogram" + type: DT_FLOAT + } + input_arg { + name: "sample_rate" + type: DT_INT32 + } + output_arg { + name: "output" + type: DT_FLOAT + } + attr { + name: "upper_frequency_limit" + type: "float" + default_value { + f: 4000 + } + } + attr { + name: "lower_frequency_limit" + type: "float" + default_value { + f: 20 + } + } + attr { + name: "filterbank_channel_count" + type: "int" + default_value { + i: 40 + } + } + attr { + name: "dct_coefficient_count" + type: "int" + default_value { + i: 13 + } + } +} +op { name: "Min" input_arg { name: "input" diff --git a/tensorflow/core/ops/ops.pbtxt b/tensorflow/core/ops/ops.pbtxt index 9725b6538d..6d28cb7e84 100644 --- a/tensorflow/core/ops/ops.pbtxt +++ b/tensorflow/core/ops/ops.pbtxt @@ -11540,6 +11540,57 @@ op { description: "result is one logical checkpoint, with one physical metadata file and renamed\ndata files.\n\nIntended for \"grouping\" multiple checkpoints in a sharded checkpoint setup.\n\nIf delete_old_dirs is true, attempts to delete recursively the dirname of each\npath in the input checkpoint_prefixes. This is useful when those paths are non\nuser-facing temporary locations." } op { + name: "Mfcc" + input_arg { + name: "spectrogram" + description: "Typically produced by the Spectrogram op, with magnitude_squared\nset to true." + type: DT_FLOAT + } + input_arg { + name: "sample_rate" + description: "How many samples per second the source audio used." + type: DT_INT32 + } + output_arg { + name: "output" + type: DT_FLOAT + } + attr { + name: "upper_frequency_limit" + type: "float" + default_value { + f: 4000 + } + description: "The highest frequency to use when calculating the\nceptstrum." + } + attr { + name: "lower_frequency_limit" + type: "float" + default_value { + f: 20 + } + description: "The lowest frequency to use when calculating the\nceptstrum." + } + attr { + name: "filterbank_channel_count" + type: "int" + default_value { + i: 40 + } + description: "Resolution of the Mel bank used internally." + } + attr { + name: "dct_coefficient_count" + type: "int" + default_value { + i: 13 + } + description: "How many output channels to produce per time slice." + } + summary: "Transforms a spectrogram into a form that\'s useful for speech recognition." + description: "Mel Frequency Cepstral Coefficients are a way of representing audio data that\'s\nbeen effective as an input feature for machine learning. They are created by\ntaking the spectrum of a spectrogram (a \'cepstrum\'), and discarding some of the\nhigher frequencies that are less significant to the human ear. They have a long\nhistory in the speech recognition world, and https://en.wikipedia.org/wiki/Mel-frequency_cepstrum\nis a good resource to learn more." +} +op { name: "Min" input_arg { name: "input" |