Abstract
The legend of Speech Recognition has long and honored history. This article discusses some key milestones of that history in both technical and commercial aspects. And also some special researches carried out and their successes and failures are discusses. Further illustrate how and who’s highly contributed to keep this technology in current position.
The original source of the image hosted at [http://www.lumenvox.com/resources/tips/historyOfSpeechRecognition.aspx]
The legend of Speech Recognition Technology has long history. It was started mid-1870s, by Alexander Graham Bell as an attempt to create speech recognition machine that would be able to transform verbal words into text (understandable picture) in a way that even a deaf person could interpret the verbal meanings. But he ended up with another machine call ‘telephone.’ [7]
Engineers, language experts and various other scientists have been attempting to make Bell’s dream a reality with some extensions. It was creating machines those capable in understanding voice of human beings. In 1920 “Radio Rex” has recorded as the earliest succeeded attempt of speech recognition technology history. It was a simple toy-a dog capable of moving when its name was called. In 1930s U.S. Government (first and foremost by the Military and DARPA - Defense Advanced Research Project Agency) and some Universities as well funded on speech recognition researches. It was a real kick off to this area and made many advances.
With this inspiration in 1936, AT&T's Bell Labs started their researches in “Voder” a manually controlled speech synthesizer and was demonstrated in 1939 at World’s Fair.
Then in 1952 a small vocabulary discreet digit recognizer (i.e. the numbers zero up to nine) was demonstrated over the telephone by the same company. Within the same decade found that there is evidence for statistical patterns in speech. It was the times that statistical model for speech recognition was introduced.
[1, 7, 8, 9]
During the next decade early digit recognizers improved. Researchers pay attention on smaller projects and initially attempted to develop desecrate speech recognition system. In this decade recognition systems migrated to acoustic model; rather make use of statistical models. Even in 1964 Martin deployed neural networks for phoneme recognition it seamed natured slowly and later replaced by new ones. Some improvements in vocabulary and accuracy can be identified. These acoustic models consider about the sonic of speech rather than the meaning and supported to implement discreet speech recognition systems well. They are used productively in modern speech recognition products, as well because their ability of working properly in noisy environments and for smaller vocabulary systems. [7]
Again the US military was funded for speech projects in 1970s. There was a project funded them to achieve the expectation of to develop continuous speech recognition system which can be operate on 90% of accuracy. And such systems are still being developed and refined today. In this decade also a new model was upped and introduced by Carnegie-Mellon and Princeton Universities to speech recognition that is "Hidden Markov Model" (HMM) which is widely use in current products as well. [7]
US military funds were continuously flooded towards speech recognition researches during this decade as well. With this helping hand a large Speech Recognition research project was conducted called CMU. Later some excellent students, those who worked in CMU, become members of Microsoft's speech recognition group which is one of most succeeded speech recognition teams. Ending up the starvation of commercial products in 1982 “Covox” released their first speech recognition product and recorded as the first of amongst all other commercial speech recognition products. A new company named Dragon Systems which is a dominant player of speech recognition also established and followed “Covox” and produced a SR system ran on personal computers. It was a simple command and control system. On the other hand ‘Bakers” produced the first, non-commercial, version of speech recognition.
Many further studies were carried out in next decade. In 1995, Dragon Systems introduced a general-purpose speech recognition product allowed users to dictate into their PC. This system expected keep pauses between each spoken word. This product is the earliest of the type (discreet speech recognition systems). The days of this product was released by Dragon, IBM also released a product. Two years later in 1997, Dragon released the first general purpose continuous speech dictation systems, named “Naturally Speaking”. It allows user to speak in conventional way of speaking to their computer as a option to pointing and to, typing. Microsoft also founded on speech recognition and started their own lab recruiting some excellent students from the CMU project, which was won the US military’s praises twice. When the decade was ended and Microsoft entering to the market Dragon systems purchased Lernout & Haupsie as well and still considered as the biggest company in whole over the world related to the speech recognition valued at $460 million at the end of 90s. In these days number of new products scattered over various domains were introduced. Amongst them call counters; the stock marker broker systems which enable customers to quotes on stocks and options over the telephone and many other telephony systems were highlighted. Also here should mention the improvements of the mobile devices mainly developed with the helping hand of SR technology. Mobile phones were manufactured in a way that user could attached voice tags on their contacts and later when needed to dial to that contact just needed to speak that tagged word called (voice dialing). Command and Control was also introduced to mobile phones enabling functionality such as turn to the phone state to silent mood just commanding to the phone using voice.
Reference:
[1] Speech Recognition Software and Medical Transcription History, A Timeline of Speech Recognition, http://www.dragon-medical-transcription.com/historyspeechrecognitiontimeline.html, 11 july 2007
[4] Histry of Speech Recognition, http://www.lumenvox.com/resources/tips/historyOfSpeechRecognition.aspx, 11 july 2007
[7] Cell Phones: History of Speech Recognition, http://www.thecellphoneforum.com/content.php.article.62, 14 July 2007
[8] Milestones in Speech Technology – Past and Future!, http://www.speechtechmag.com/Articles/ReadArticle.aspx?ArticleID=29572, 14 July 2007
[9] Speech Recognition, Daniel Thalman, http://209.85.129.104/search?q=cache:04GGydLYegYJ:vrlab.epfl.ch/~thalmann/VR/VRcourse_Speechrec.pdf+History+of+speech+recognition&hl=en&ct=clnk&cd=32, 14 July 2007
[10] Speech Recognition Research Paper.
