【摘要】：論文中將介紹一種經(jīng)過(guò)優(yōu)化的語(yǔ)音合成以及分發(fā)系統(tǒng)的方案。與原有的系統(tǒng)相比,新型的語(yǔ)音系統(tǒng)采用分布式設(shè)計(jì),通過(guò)優(yōu)化系統(tǒng)結(jié)構(gòu),服務(wù)器／客戶端的緩存分發(fā)方法,使得新型系統(tǒng)在可靠性,實(shí)時(shí)性,以及實(shí)施成本上,都有了令人振奮的表現(xiàn)。目前大部分智能語(yǔ)音廣播系統(tǒng)中,一般都會(huì)包括語(yǔ)音服務(wù)器,音頻矩陣以及功放這三個(gè)主要部件。其中,服務(wù)器擔(dān)負(fù)著接收處理外部接口信息,語(yǔ)音片的匹配合成,音頻矩陣的路由切換,以及語(yǔ)音播放等諸多功能,由于技術(shù)問(wèn)題以及一些歷史問(wèn)題,使得服務(wù)器成為了整個(gè)語(yǔ)音系統(tǒng)中最主要的性能瓶頸。從軟件技術(shù)的角度來(lái)講,如此設(shè)計(jì)的會(huì)帶來(lái)系統(tǒng)響應(yīng)緩慢,系統(tǒng)資源管理不便等問(wèn)題；從系統(tǒng)部署的角度來(lái)講,整個(gè)系統(tǒng)難以模塊化,難以升級(jí)。新型的語(yǔ)音系統(tǒng)中,原有“單獨(dú)”的服務(wù)器將會(huì)設(shè)計(jì)成分布式服務(wù)器／客戶端的形式,但是客戶端的概念將被弱化,客戶端相對(duì)于服務(wù)器而言,成為服務(wù)器的子節(jié)點(diǎn)。其中,服務(wù)器只負(fù)責(zé)語(yǔ)音合成以及分發(fā),客戶端(子節(jié)點(diǎn))負(fù)責(zé)切換路由與語(yǔ)音播放。由于采用分布式設(shè)計(jì),在整個(gè)智能語(yǔ)音系統(tǒng)中,可依據(jù)現(xiàn)場(chǎng)環(huán)境規(guī)模,部署多個(gè)服務(wù)器和客戶端。分布式設(shè)計(jì)將會(huì)分擔(dān)原有“單獨(dú)”服務(wù)器的系統(tǒng)負(fù)載,從而使整個(gè)系統(tǒng)在實(shí)時(shí)性上大幅提高,并且將原有“集權(quán)”式的資源管理模式分?jǐn)偟礁鱾�(gè)模塊中,此舉將大大便于資源管理,從而降低系統(tǒng)出錯(cuò)率。另外,分布式設(shè)計(jì)使得系統(tǒng)部署靈活,各模塊間可以自由組合,可依據(jù)成本按須訂制,從而在產(chǎn)品推廣上會(huì)成為一大優(yōu)勢(shì)。
[Abstract]:In this paper, an optimized scheme of speech synthesis and distribution system is introduced. Compared with the original system, the new voice system adopts the distributed design. By optimizing the system structure and the buffer distribution method of the server / client, the new system can be implemented in terms of reliability, real-time performance and implementation cost. There are exciting performances. At present, most intelligent voice broadcasting systems generally include three main components: voice server, audio matrix and power amplifier. Among them, the server is responsible for receiving and processing external interface information, speech chip matching synthesis, audio matrix routing, voice playing and many other functions, due to technical problems and some historical problems, The server has become the main performance bottleneck in the whole voice system. From the point of view of software technology, the system response is slow and the system resource management is inconvenient. From the point of view of system deployment, the whole system is difficult to modularize and upgrade. In a new type of voice system, the original "separate" server will be designed as a distributed server / client, but the concept of client will be weakened, and the client will become a sub-node of the server relative to the server. The server is only responsible for speech synthesis and distribution, and the client (sub-node) is responsible for switching routing and voice playing. Because of the distributed design, multiple servers and clients can be deployed in the whole intelligent voice system according to the scale of the field environment. The distributed design will share the system load of the original "separate" server, so that the whole system will be greatly improved in real-time, and the original "centralized" mode of resource management will be distributed to each module. This will greatly facilitate resource management, thereby reducing the system error rate. In addition, distributed design makes the system deployment flexible, each module can be freely combined, can be customized according to cost, so it will become a great advantage in product promotion.
【學(xué)位授予單位】：復(fù)旦大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：TP311.52;TN912.3

【參考文獻(xiàn)】

相關(guān)期刊論文 前3條

1 桂盛霖;羅蕾;李允;于淼;徐建華;;基于自動(dòng)機(jī)理論的分布式實(shí)時(shí)調(diào)度分析工具[J];軟件學(xué)報(bào);2011年06期

2 楊戈;樊秀梅;;基于CDN的流媒體動(dòng)態(tài)調(diào)度算法[J];通信學(xué)報(bào);2009年02期

3 郝婕;;通訊系統(tǒng)中語(yǔ)音信號(hào)合成算法研究[J];浙江萬(wàn)里學(xué)院學(xué)報(bào);2006年02期

，

本文編號(hào)：2387005