Wayne

when use DBiRNN and lstm. In utils, line 210 shape = [len(targetList), np.asarray(indices).max(0)[1]+1] #shape

If we know the character ,why not use rnn? Thanks, your work is very excellent.

tensorflow test error as_list() is not defined on an unknow tensorshape

2

ERROR: test_multiple_batches_cpu (test_warpctc_op.WarpCTCTest) ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/work/speech/libs/warp-ctc/tensorflow_binding/tests/test_warpctc_op.py", line 124, in test_m ultiple_batches_cpu self._test_multiple_batches(use_gpu=False) File "/home/work/speech/libs/warp-ctc/tensorflow_binding/tests/test_warpctc_op.py", line 121, in _test_ multiple_batches use_gpu=use_gpu) File "/home/work/speech/libs/warp-ctc/tensorflow_binding/tests/test_warpctc_op.py", line 28, in...

请问您的模型测试集上的phone error rate是多少呢

3

谢谢

why? How about you ?3x

Try this architecture in the speech enhancement for speech recognition?

5

Can this architecture recover speech from the noisy speech?

Is there any plan to maintain the website?