Abstract: The time\|sequence pattern is a pattern,whose characteristic space is distributed on the time\|axis,such as speech signal,radar signal,etc..Presented in the paper is an improved recurrent neural network methodthe time\|tag recurrent neural network （TTRNN） method,to classify the time\|sequence patterns directly,which overcomes the drawbacks of the traditional method,and a good result is achieved.The preliminary experiment not only proves the good performance of the TTRNN method in the clasification of the time\|sequence patterns,but also proves the importance of time\|tag in the classification of the time\|sequence patterns.