Meet tensorflow half a month ago, want to write a own speech recognition program, when I use thchs30 after training corpus, but couldn't find how to call their own recorded voice, come to help you a great god, and hope can solve the younger brother for half a month of doubt,,,,

Below is test training model code, but I don't know how to input custom audio identification:

 
The import OS 
The import difflib 
The import tensorflow as tf 
The import numpy as np 
The from utils import decode_ctc, GetEditDistance 


# 0. Ready to decode the dictionary, parameters should be in line and training, and can also be saved to the local dictionary, direct read 
The from utils import get_data, data_hparams 
Data_args=data_hparams () 
Train_data=https://bbs.csdn.net/topics/get_data (data_args) 


Acoustic model # 1. -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
The from model_speech. Cnn_ctc import Am, am_hparams 

Am_args=am_hparams () 
Am_args. Vocab_size=len (train_data. Am_vocab) 
Am=am (am_args) 
Print (' loading & model... ') 
Am. Ctc_model. Load_weights (' logs_am/model. The h5) 

# 2. The language model -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
The from model_language. Transformer import Lm, lm_hparams 

Lm_args=lm_hparams () 
Lm_args. Input_vocab_size=len (train_data. Pny_vocab) 
Lm_args. Label_vocab_size=len (train_data. Han_vocab) 
Lm_args. Dropout_rate=0. 
Print (' loading language model... ') 
Lm=lm (lm_args) 
Sess=tf. The Session (graph=lm. Graph) 
With lm. Graph. As_default () : 
Saver=tf. Train. Saver () 
With sess. As_default () : 
Latest=tf. Train. Latest_checkpoint (' logs_lm) 
Saver. Restore (sess, latest) 

# 3. Prepare test data needed, don't have to agree and training data, by setting the data_args. Data_type test, 
# should be set to 'test' here, because I use the 'train' presentation model is small, if you use the 'test' can't see effect, 
# and there will be a word does not appear, 
Data_args. Data_type='test' 
Data_args. Shuffle=False 
Data_args. Batch_size=1 
Test_data=https://bbs.csdn.net/topics/get_data (data_args) 

Test # 4. -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
Am_batch=test_data. Get_am_batch () 
Word_num=0 
Word_error_num=0 
For I in range (10) : 
Print (' \ n the 'I,' th example.) 
# load the trained model, which can identify and 
Inputs, _=next (am_batch) 
X=inputs [' the_inputs] 
Y=test_data. Pny_lst [I] 
Result=am. Model. Predict (x, steps=1) 
# will be digital results into text 
_, text=decode_ctc (result, train_data am_vocab) 
Text="'. Join (text) 
Results: print (' text ', text) 
Print (' the original result: ', ' 'the join (y)) 
With sess. As_default () : 
Text=text. The strip (' \ n '). The split (' ') 
X=np. Array ([train_data. Pny_vocab. Index (pny) for pny in text]) 
X=x.r eshape (1, 1) 
Preds=sess. Run (lm) preds, {lm. X: x}) 
Label=test_data. Han_lst [I] 
Got="'. Join (train_data han_vocab [independence idx] for independence idx in preds [0]) 
Print (' the original characters: 'label) 
Print (' recognition results: ', got) 
Word_error_num +=min (len (label), GetEditDistance (label, got)) 
Word_num +=len (label) 
Print (' word error rate: 'word_error_num/word_num) 
Sess. Close () 

CodePudding user response:

Oh ~ how no one?

CodePudding user response:

A ~ ~ ~ ~

CodePudding user response:

Everyone a great god ~ please help ~ ~ ~