Tools
Tools created during the course of the EMIME project and made available open-source. (Use option Add new -> link to add more)
-
Java HMM editor
- This is a program written in java to perform decision tree marginalisation operations on HTK HMM files.
-
Cross language state mapping
- A set of programs written mainly in perl to do cross language state mapping and adaptation.
-
WSJ training scripts
- A set of scripts to train ASR models. They are based on HTS scripts from Junichi Yamagishi, but modified to build ASR models. They were used for the experiments in the "Measuring the gap" paper.
-
VTLN as CSMAPLR prior
- This code will allow VTLN to be used in combination with CMLLR adaptation by using a VTLN transformation matrix as a prior to the CSMAPLR transformation. This technique can be a global VTLN prior matrix in the structural MAP adaptation or multiple VTLN matrices as priors to the different nodes in the regression tree during CMLLR adaptation.
-
EM VTLN adaptation
- This allows VTLN adaptation within the context of HTS. This code will include the possibility to perform VTLN adaptation as a global warping of the spectrum using base classes and also as multiple warping parameters for different phoneme classes using regression trees (similar to CMLLR adaptation). The code will be incorporated into the mainline HTS distribution (HTS 2.2 beta) from NITech.
-
Framework of EMIME Research Demo
- EMIME Research Demo is for component and whole-system evaluation in Speech to Speech translation. The package maintains the framework of EMIME Research Demo. In this public released package, a few modules are omitted but you can easily fulfill it following README.
-
Framework of EMIME Realtime Demo
- EMIME Realtime Demo is a real-time embedded demonstrator for user evaluations in realistic scenarios of Speech to Speech translation. The package maintains the framework of EMIME Realtime Demo. In this public released package, a few modules are omitted but you can easily fulfill it following README.
-
Client Software of EMIME Realtime Demo
- Client software of the EMIME Realtime Demo This client software fits for smart phones based on the Symbian OS S60 5th edition platform.
-
HMM Training and adaptation scripts for CSTR/EMIME HTS-2010 TTS systems
- HTS voice building scripts that support parallel processing. This was used for building the CSTR/EMIME HTS 2010 system.
-
Autoregressive HMM for HTS
- An implementation of the autoregressive HMM built on top of HTS. It provides the ability to do embedded re-estimation and decision tree clustering for the autoregressive HMM.
-
Autoregressive HMM version of the HTS demo
- An autoregressive HMM version of the HTS demo. Provides an example of how to use the Autoregressive HMM for HTS patch above.
-
armspeech
- Flexible python framework for probabilistic modelling of speech with a focus on autoregressive models. Allows the use of more complicated autoregressive-style output distributions than the HTS implementation above. Includes example experiments.