Emotional speech database is the foundation of emotional speech recognition research,it has great significance to establish a continuous dimension emotional speech database including cognitive psychological factors for improving the performance of the speech emotion recognition and human-computer interaction.In this paper,first,hearing screening was conducted on previously established TYUT2.0 database,then the database was optimized according to recognition rate threshold.The resultant emotional speech database with 237 speeches has four types of emotion including 62,58,57,and 60 speeches representing respectively sadness,anger,happiness and surprise.The speech of this database was marked by using PAD emotion model,giving a dimensional emotion database.Each speech has its identification rate and PAD value.Statistical results of PAD value prove the validity of this dimensional emotional speech database,which lays the foundation for studying emotional speech recognition in continuous dimension in the future.