BaiduCN1.2k Model识别结果不对
Created by: harold-yh
使用deploy_server.py和deploy_client.py进行识别,声音来源麦克风,音频文件正常,识别结果不对 语言模型使用:Mandarin LM Small Speech Model使用:BaiduCN1.2k Model,参数按照BaiduCN1.2k Model的README设置,识别出来的结果完全不对,一句话只识别成一个字,并且还是错的,麻烦指导一下是我使用的不对吗?
----------- Configuration Arguments -----------
alpha: 1.15
beam_size: 500
beta: 0.15
cutoff_prob: 1.0
cutoff_top_n: 40
decoding_method: ctc_beam_search
host_ip: localhost
host_port: 8086
lang_model_path: models/lm/MandarinLMSmall.klm
mean_std_path: models/baidu_cn1.2k_model/mean_std.npz
model_path: models/baidu_cn1.2k_model/params.tar.gz
num_conv_layers: 2
num_rnn_layers: 3
rnn_layer_size: 2048
share_rnn_weights: 0
specgram_type: linear
speech_save_dir: demo_cache
use_gpu: 0
use_gru: 1
vocab_path: models/baidu_cn1.2k_model/vocab.txt
------------------------------------------------
结果全是下面这样的 Recognition Results: 人 Recognition Results: 在 Recognition Results: 我