当下语音识别技术.docxVIP

Speech Recognition: State of the artAC41003 mini-projectXueying WuBackgroundSpeech recognition, which is also known as automatic speech recognition (ASR), is used to transfer human being’s voice into computer understandable data which then be turned into texts or commands. Main ASR usage includes: document dictation, PC navigation and control, telephone-based enquiry.The probably earliest speech recognition gadgetry is a toy dog named “Radio Rex” which can pop out of its kennel when its name is called [1]. The first speech recognition system Audrey which is based on computer is developed by ATT which can understand 10 digits with an accuracy of 98% [2]. Followed by Linear Predictive Coding (LPC) and Dynamic Time Warp technology, speech recognition technology reached a higher stage. The biggest breakthrough in speech recognition is Hidden Markov Model (HMM): a well known and widely used statistical method of characterizing the spectral properties of the frames of a pattern. The HMM method provides a natural and highly reliable way of recognizing speech for a wide range of applications and integrate well into systems incorporating both task syntax and semantics [3]. Nowadays, the basic technology applied in speech recognition is still more or less relying on HMM. Artificial neural network (ANN) is another model that is researched in speech recognition field. However, ANN does not have the ability of dynamic time reforming which is used in HMM, while HMM is not capable of applying functional aspects of biological neural networks [4]. The current preferred use is to make HMM to identify speech sounds while use ANN to store sounds.Speech recognition technology has been extensively studied during the past few decades. The existence of commercial products is more than 20 years, at first for isolated word recognition, and then for connected words and continuous speech [5]. Most of these systems are based on statistical modelling, both at the acoustic and linguistic levels [