Continuous voice recognition software or speech recognition software (also known as automatic speech recognition, computer speech recognition, speech to text, or just STT) converts spoken words into text. The term 'voice recognition' is used to refer to systems that must be trained to a particular speaker and are commonly used in healthcare settings to replace or improve the efficiency of medical transcribers (usually medical secretaries). Early systems required the speaker to pause between each word, were slow, and had limited vocabulary and high error rates. Recent advances in this field have generated newer systems that understand continuous speech, run on common personal computers and produce more accurate results. 1Voice recognition systems comprise a microphone that converts speech to an analogue electrical signal, which is converted to a digital signal by an electronic circuit board within a computer. 2 Speech recognition engine software then uses acoustic, language and vocabulary models as well as complex statistical algorithms to transform the digital signal into words and punctuation marks. The acoustic model removes noise and unnecessary information such as changes in volume. The language model then analyses the content of the speech; it compares the combinations of phonemes with the words in its digital dictionary, a huge database of the most common words in the English language. Most of today's packages come with dictionaries containing about 150 000 words. The language model quickly decides which words were said and displays them on the screen. The commercial software package used for this study is widely used in UK healthcare within National Health Service (NHS) trusts and general practices to produce clinical correspondence using digital dictation and transcription (similar software packages are available in other countries). It includes a wide selection of medical terms and comes with a 'training wizard' which learns new words and also adapts itself to the voice of a new user. The software has wide applications in commercial settings, for example in automated telephone messaging services, which have a limited vocabulary for a wide range of users; other applications of voice recognition may have a large vocabulary trained to work best with a small number of users, such as in digital transcription services. Voice recognition in healthcareThere is a huge pressure in healthcare settings to generate large amounts of documentation in a short time. The use of computerised voice recognition in medicine was first described in radiology in 1981.3 In early studies the higher error rate of digital transcription when compared with traditional typing was highlighted. The main reasons for the high error rate then were the necessity to speak each word distinctly in a monotonous voice in order for the computer to recognise it. This also resulted in the doctor spending more time in correcting errors as they appeared on the screen in real time. Things have changed over the years in that newer software is able to r...