Speech Recognition

Speech Recognition Technology is not a strange technology and it has a long and venerable history. The idea of making machines will capable in responding for human commands, was exits from many time ago because it seams a real option for user input. It carries the potential of reducing the existence distance between machines and human beans and currently has achieved a high level of accuracy (Over 95%) and performance. Also the industry and research of speech recognition is very active, evolving and changing rapidly resulting that a better place for not only regular computer users but also for physically handicapped individuals.

Another ability of Speech recognition is transforming spoken words into text. The impotence of this is though Speech is easier to generate, more conventional and fast in generation. On the other hand writing text is slow and a hard process like in a case of interviewing. But most of the times people like to and required to have text versions of these speeches. It may cause of listening to speech is slower, harder to memorize, harder to navigate and harder revise and many other practical issues. Converting of speech to text can makes a better solution to these issues.

Even though the speech recognition idea and some technologies were exists for years the progress of the technology was staged due to lack of hardware availability in past. But as a result of development of computer processing power and their availability at cheap prices also with the helping hands of many enthusiastic researchers, sponsors and many other stakeholders the speech recognition technology has been developed and still is developing rapidly. Today, there are diverse of applications scattered over number of domains based on speech recognition such as Dictation systems, as products developed for general computer interaction (Voice enable interfaces), IVR systems, Voice enable web sites, Language translators and many more.

Actually though these application are often considered as simply speech recognition systems most of them are hybrid systems of both Speech-Recognition and Voice-Recognition where Speech-Recognition is simply a process that identify spoken words and Voice-Recognition is a process in which identifying the producer of spoken words.

Modern speech recognition systems are developed to one of two main categories of customer requirement\specifications. All those systems have been ended upped as either a speaker-dependent continuous-speech PC-based system or a speaker-independent continuous-speech server-based system.

Speaker independent systems are generally expensive than individual speaker-dependent system and don’t expect previous training for particular user. Even such systems are expensive when deal with huge and various user bases it is require to have speech independence and performance rather when deal with one or two single subscribers.

There is no doubt of these statements and facts. All of them are true for English language speech recognition but not sure for non English languages. Even though most of these systems has developed for English language, researches has realized that the requirement of having those for non English languages as well. Number of projects and researches has been accomplished for non English languages as well but this numbers are very smaller when it compares with projects has been conducted for English language speech recognition.

References:

[1] White Paper On Speech Recognition In The SESA Call Center, By Ron Mains, Tim Meier, Scott Nainis, Henry M. James,http://www.itsc.state.md.us/PDF/O-2-2%20Technology%20Assessment%20Final%20Report.pdf, April 2001.

[4] Histry of Speech Recognition, http://www.lumenvox.com/resources/tips/historyOfSpeechRecognition.aspx, 11 july 2007