本人尝试利用BNT工具包，进行语音识别相关研究。模型采用DBN，学习算法为EM，推理采用联合树，但在提取语音特征时，出现如下故障如下：每两次迭代表是一个说话人的特征参数，但是12个人的特征似然值是相同的，请问这是为什么？求各位大侠指教！
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
EM iteration 1, loglik = -360436.5339
EM iteration 2, loglik = -360436.5339
