论文部分内容阅读
为了提高电子商务中用户认证的安全性,提出了一种高识别率的判别最大熵语音识别机制DME.该语音识别方法同时考虑语音与语言两方面的因素,并将语音和语言特征进行有效地结合,在统一的最大熵模型下,实现判别训练,确保观察样本能正确地分配到其对应状态,以提高所训练出的语音模型的正确识别率.详细的实验及与现有方法的比较结果表明,对不同环境下的语音数据,提出的语音识别方法具有更好的识别性能,对提高电子商务中用户认证系统的安全性能具有理论与实际意义.
In order to improve the security of user authentication in e-commerce, a high recognition rate discriminant maximum entropy speech recognition mechanism DME is proposed. The speech recognition method considers both speech and language factors, and effectively combines the speech and language features Combined with the unified maximum entropy model to achieve discriminant training to ensure that the observed samples can be correctly assigned to its corresponding state in order to improve the correct recognition rate of the trained voice model.Detailed experiments and comparison with existing methods It shows that the proposed speech recognition method has better recognition performance for speech data under different environments and has theoretical and practical significance to improve the security performance of user authentication system in e-commerce.