摘要
采用按需分段合成的策略,即服务器每次只合成一小段文本给客户端使用,当客户端播放语音数据时服务器再合成另一小段文本,以提高在单机情况下系统支持的并发线路数;再根据负载均衡的原理使用多台计算机并行处理文语转换运算.对语音合成服务器的性能测试结果表明,采用按需分段合成策略和负载均衡的设计可提升文语转换系统所能达到的规模,并能够满足系统的实时性、并发性、稳定性和系统无级扩容的要求.
Presents a strategy that synthesizes a bit of text for the client port each time, when the client port plays back the speech data, the speech synthesizing server port synthesizes the next bit of text to improve the concurrent lines of a single computer, and then utilizes many computers to synthesize speech concurrently according to the principle of load balance. Test results of performance about speech synthesizing server indicate that the system can synthesize speech in a large scale through subsection synthesis and load balancing, and satisfy the request of real-time, concurrency, stability and extension, through the subsection synthesis method and load balance design.
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2004年第1期74-76,81,共4页
Transactions of Beijing Institute of Technology
关键词
文语转换
分段合成
负载平衡
并发
text to speech (TTS)
subsection synthesis
load balancing
concurrency
