TOKUDA, NANKAKU and HASHIMOTO LABORATORY - HOME/SOFTWARE の履歴(No.9)

SOFTWARE

In Tokuda & Lee & Nankaku laboratory, softwares for promotion of speech and image research are developed and opened to the public. These are used to research in various organizations and companies.

Software

HMM Speech Sysnthesis System Toolkit: HTS

HTS is a basic software for speech synthesis that a lot of research laboratories (Microsoft, IBM, etc) adopt.

Open-Source Large Vocabulary Continuous Speech Recognition Engine: Julius

Julius is an open-source and a high-performance, general-purpose large vocabulary continuous speech recognition engine for developments and researches of the speech recognition system. It can perform almost real-time decording on most current PCs in tens of thousands of vocabulary. It has high generality and can apply to various, wide usages by rearranging modules such as a pronunciation dictionary, a language model and an acoustic model, etc. The function is offered in the library, and the inclusion to application is also possible.

Speech Signal Processing Toolkit: SPTK

This is a software that does signal processing and data processing for the acoustical analysis.

Speech Synthesis Engine: hts_engine

It is a software which synthesis speech by using model learned by HTS. It opens to the public with BSD license.

Japanese Text-to-Speech System: Open JTalk

It is a Japanese Text-to-Speech System. hts_engine is used in a speech synthesis module. It opens to the public with BSD license.

Voice Interaction System Building Toolkit: MMDAgent

MMDAgent is an open-source, toolkit for building voice interaction systems.
This realizes to talk with characters (MMD model) in the screen by combining speech synthesis, speech recognition, speech learning function, 3D rendering, lip sync technology, etc.

Anthropomorphic Spoken Dialogue Agent: Galatea

This is an open-source, license-free software toolkit for building anthropomorphic spoken dialogue agents. This is a product of project which speech, language, and image researchers from ten or more university in Japan participate to build anthropomorphic spoken dialogue agents. HTS and Julius, those are developed in this laboratory, are used on speech wave form generation module and speech recognition module respectively in this software.

System

Nitech Campus Information Guidance System "Meichan" (Japanese page)

Speech Information Guidance System is located at the ground floor of second building in Nagoya Institute of Technology. Please talk when you come Nagoya Institute of Technology.

Database

Multi-Modal Speech Database for Research: M2TINIT (Japanese page)

M2TINIT is a multi-modal database which japanese speech and lip dynamic scene are recorded concurrently. It is developed and opened to the public by Takao Kobayashi laboratory (Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology) and Kitamura & Tokuda laboratory (Department of Computer Science, Nagoya Institute of Technology. Currently, Tokuda & Lee & Nankaku laboratory) for promotion of multi-modal speech research. It has been used to researches that are generation of speech and lip dynamic scene and bimodal speech recognition.