reading MFCC features from CSV files instead of extracted them on the fly.

[kerolos]

trafficstars

In this Tutorial (keras-io/examples/audio/ctc_asr.py) https://keras.io/examples/audio/ctc_asr/#:~:text=CTC%20is%20an%20algorithm%20used,create%20is%20similar%20to%20DeepSpeech2.

The features had been extracted from wav files path with is in dataframe : def encode_single_sample(wav_file, label): file = tf.io.read_file(wavs_path + wav_file + ".wav")

Is there any way to read the features MFCC from CSV files in this function feature_dim=41 font_column_types= [float()]*(feature_dim) numeric_features = tf.io.decode_csv(feature_file,record_defaults=font_column_types)