Application Study Of Personalized Speech Synthesis On Ship Navigation System

With the rapid development of navigation technology and computer technology, the automation degree of ship navigation becomes higher and higher, but most of the ship navigation system’s alarms are broadcasted with mechanical voice, seafarers need to obtain the alarms with the naked eyes.To a certain extent, it affects the safety of navigation. In recent years, speech recognition technology has been applied successfully in the field of navigation, it is promising that speech synthesis technology is applied to navigation. This paper is doing this attempt. The main purpose of this paper is to replace the mechanical voice of the ship navigation system alerts with natural speech which is the target person’s voice, such as the captain’s voice. This can make the ship navigation system alarms more authoritative.Take Hidden Markov Model based statistical parameters speech synthesis as the foundation of research, with the speaker adaptive method, by using four speakers’ corpus and one target speaker corpus, this paper realizes the personalized speech synthesis, and successfully applies to the ECDIS system. The main works of this paper are summarized as follows:(1) Establish a VW Pual target speaker’s corpus, including collecting audio files and the front-end text analysis.(2) Compile the tools on Linux system, such as HTK, HTS, SPTK, Festival, STRAIGHT, etc. That are needed in speech synthesis system. Realize SD(Speaker Dependent) speech synthesis, which is based on a single corpus.(3) Realize the personalized English speech synthesis by using the speaker adaptive technology. Firstly, do the speaker adaptive training to four speakers’ speech data, and get the average voice model. Then using the CMLLR adaptive algorithm to obtain the speech which is similar to the target speaker’s speech. Finally, the experiment test results show that compared with the voice synthesised by using SD speech synthesis, the voice synthesised by using the speaker adaptive speech synthetic is more similar to the target speaker VW Pual.(4) Realize sixteen alarm detecting of ECDIS by programming with Visual Studio 2010. And finally realize the personalized speech alarm on the electronic chart display system.