System Overview Front-end Parameterization Acoustic 0 0 0 0 0 Acoustic Modelling Language Modelling Tahle 1: Iltk Lvr Performance Oil Arpa Csr I %word Vocab Size 1 Lm N-gram J Adapted Error Current State of Lvr Current Issues Real-time Operation

progress has been made in speech-recognition technology over the last few years and nowhere has this progress been more evident than in the area of large-vocabulary recognition (LVR). Current laboratory systems are capable of transcribing continuous speech from any speaker with average word-error rates between 5% and 10%. If speaker adaptation is allowed, then after 2 or 3 minutes of speech, the error rate will drop well below 5% for dependent and required words to be spoken with a short pause between them. However, the capability to recognize natural continuous-speech input from any speaker opens up many more applications. As a result, LVR technology appears to be on the brink of widespread deployment across a range of information technology (IT) systems. This article discusses the principles and architecture of current LVR systems and identifies the key issues affecting most speakers. LVR systems had been limited to dictation applications since the systems were speaker their future deployment. To illustrate the various points raised, the Cambridge University HTK system is described. This system is a modem design that gives state-of-the-art performance , and it is typical of the current generation of recognition systems. Current LVR systems are firmly based on the principles of statistical pattem recognition. The basic methods of applying these principles to the problem of speech recognition were pioneered by Baker, Jelinek, and their colleagues from IBM in the 1970s, and little has changed since [13, 541. Figure 1 illustrates the main components of an LVR system. An unknown speech waveform is converted by a front-end signal processor into a sequence of acoustic vectors, Y = y i. y2, ...,Y T. Each of these vectors is a compact representation of the short-time speech spectrum covering a period of typically 10 msecs. Thus, a typical 10-word utterance might have a duration of around 3 seconds and would be represented by a sequence of T = 300 acoustic vectors. The utterance consists of a sequence of words, W = wi, w2,..wn, and it is the job of the LVR system to determine the most probable word sequence, W , given the observed acoustic signal Y. To do this, Bayes' rule is used to decompose the required probability P(WIY) into two components, that is, This equation indicates that to find the most likely word sequence W, the word sequence that maximizes the product of P(W) and P(YIW) must be found. The first of …