Rapid CODEC adaptation for cellular phone speech recognition
نویسندگان
چکیده
Along with the ever increasing popularity of cellular phones, improving recognition accuracy in cellular phone speech has become an issue of growing concern. However, the distortion caused by current low-bit rate speech CODEC is nonlinear, so compensating for distortion by applying only a conventional CMN which assumes distortion is a stationary linear transfer on the cepstrum domain is di cult. In this paper, to improve speech recognition accuracy over cellular phone networks, we investigate the use of CODEC-dependent acoustic models and rapid CODEC adaptation using model selection based on maximum likelihood criterion. By using these methods we succeeded in reducing recognition errors in cellular phone speech by 33 %.
منابع مشابه
A Noise Robust Voice Input System for Internet Services over Cellular Phones
Internet access services over wireless networks have already been widely used by cellular phone users in Japan [1]. However users have difficulty using keypads to browse web contents. To reduce this difficulty and offer smooth Internet access via mobile web browser, we have developed a voice input system that works on existing Internet services over cellular phones. This system works based on a...
متن کاملSpeaker Adaptation in Continuous Speech Recognition Using MLLR-Based MAP Estimation
A variety of methods are used for speaker adaptation in speech recognition. In some techniques, such as MAP estimation, only the models with available training data are updated. Hence, large amounts of training data are required in order to have significant recognition improvements. In some others, such as MLLR, where several general transformations are applied to model clusters, the results ar...
متن کاملSpeaker Adaptation in Continuous Speech Recognition Using MLLR-Based MAP Estimation
A variety of methods are used for speaker adaptation in speech recognition. In some techniques, such as MAP estimation, only the models with available training data are updated. Hence, large amounts of training data are required in order to have significant recognition improvements. In some others, such as MLLR, where several general transformations are applied to model clusters, the results ar...
متن کاملPersian Phone Recognition Using Acoustic Landmarks and Neural Network-based variability compensation methods
Speech recognition is a subfield of artificial intelligence that develops technologies to convert speech utterance into transcription. So far, various methods such as hidden Markov models and artificial neural networks have been used to develop speech recognition systems. In most of these systems, the speech signal frames are processed uniformly, while the information is not evenly distributed ...
متن کاملAcoustic model adapatation for coded speech using synthetic speech
In this paper, we describe a novel acoustic model adaptation technique which generates “speaker-independent” HMM for the target environment. Recently, personal digital assistants like cellular phones are shifting to IP terminals. The encoding-decoding process utilized for transmitting over IP networks deteriorates the quality of speech data. This deterioration causes degradation in speech recog...
متن کامل