Embedded equipment, bimodule voice synthesis system and method

Abstract

The invention provides an embedded device, a bimodal speech synthesis system and a bimodal speech synthesis method. According to the invention, the embedded device consists of a network availability detecting unit, a parameter synthesizing unit and a timbre detecting unit. The network availability detecting unit is used for detecting availability of the network. If the network is available, the received text is transmitted to a splicing and synthesizing unit at the server side by the network; if the network is unavailable, the text is input to the parameter synthesizing unit. The parameter synthesizing unit is used for parameter speech synthesis process for the entered text and outputting the processed speech to the timbre detecting unit. The timbre detecting unit is used for receiving the parameter synthesis result from the parameter synthesizing unit and the splicing and synthesizing result from the splicing and synthesizing unit at the server side within the scope permitted by realtime. Speech quality evaluation is carried out on the results so as to select the synthesizing result with the best timbre quality to be output.

Description

technical field [0001] The present invention relates to speech synthesis technology for arbitrary text-to-natural speech conversion on embedded devices, more specifically, to an embedded device, a dual-mode speech synthesis system and a dual-mode speech synthesis method, which can meet the requirements of Based on the user's requirements for real-time and sound quality, it provides users with high-quality voice synthesis output. Background technique [0002] With the advent of the digital age, voice interaction technology has been used more and more. As an important part of voice interaction, speech synthesis technology from text to speech has attracted more and more attention from academia and industry. Many companies, universities and scientific research institutes at home and abroad have done extensive and in-depth research on speech synthesis technology, and proposed waveform splicing synthesis technology based on pre-recorded speech database (see references [1][2]), spe...

AI Technical Summary

Problems solved by technology

The disadvantage is that in order to obtain better synthesis results, a large amount of voice data needs to be stored in advance, and the synthesis results are directly related to the size of the stored data, that is, when the voice library is greatly reduced, its sound quality will also be greatly reduced

According to the characteristics of embedded devices, researchers (see reference [6]) have transplanted splicing speech synthesis technology to embedded devices by simplifying the text analysis and prosody prediction modules and reducing the number of speech units in the sound bank. But what followed was a drastic drop in synthetic sound quality and naturalness

When parametric synthesis is applied in embedded devices (see reference [7]), resource occupation is not a problem, but in many cases the synthesized voice it provides does not satisfy users

Images