Model-Integration Rapid Training bas for Speech Recog
نویسنده
چکیده
Speech recognition technology has been widely used. Considering a training cost of an acoustic model, it is beneficial to reuse pre-existing acoustic models for making a suitable one for various apparatus and application. However, a complex acoustic model for high CPU power does not work for low CPU power. And a simple model for fast-processing-demanded application does not work well for high-precision-demanded ones. Therefore, it is important to adjust a model complexity according to apparatus or application, such as a number of mixture of Gaussians. This paper describes a new model-integration-type of training for obtaining a required number of mixture of Gaussians. This training can alter a number of mixture into a required one according to a specification of apparatus or application. We propose a model integration rapid training based on maximum likelihood, and evaluate the recognition performance successfully.
منابع مشابه
Rapid EM training based on model-integration
Recently, speech recognition technique has started being used in various products. In order to make a good acoustic model, usually a lot of training speech data is needed. However, due to the right of voice and privacy issues, it is not easy to collect a lot of training data. Statistical models which have been trained and transformed from speech data do not have the above mentioned problem, and...
متن کاملLikelihood Combination and Recog the Decoding of Non-native Speech
In this paper we report on the combination of multilingual Hidden Markov Models for the recognition of non-native speech. Using a digit recognition task as an example, we first demonstrate the benefits of bilingual acoustic models that incorporate training data from both the target language and the speakers’ native language, and then compare two different recognizer combination methods, namely ...
متن کاملIntegrated Multilingual Speech Recognition Impact on Chinese Spoken Language Processing
The notion of integrated multilingualism is introduced as the basis for a novel approach to multilingual speech recog nition This approach enables training of the recognizer using the data from only source language s The trained recognizer is nevertheless deployable directly to new tar get languages The performance of the recognizer is incre mentally improved via a language adaptation strategy ...
متن کاملExperiments on Chinese Speech Recog Pitch Estimation Using the M
Automatic speech recognition of a tonal and syllabic language such as Chinese Mandarin poses new challenges but also offers new opportunities. We present approaches and experimental results concerning the choice of base units for acoustic modeling, pitch estimation and how to integrate pitch estimates into the modeling framework. The experimental evaluations are carried out both on rather clean...
متن کاملIncreasing the Effectiveness of Russian Language Teaching for Special Purposes (to the Problem of Integration of Language Training with Information Technology Courses)
The article is devoted to the problem of increasing the efficiency of language teaching for the special purposes of foreign students in studying Russian at a technical university. Particular attention is paid to the training of foreign students in the skills of working with information using the latest computer technology. The conclusions of the work are based on the analysis of the results of ...
متن کامل