Looking at Speech Recognition Part 2

If you’ve never seen speech recognition demonstrated by a trained user, you’re in for a shock. A person speaks into a microphone headset and, as they speak, the words pop up on the screen. It is amazing to watch and even more amazing to use it.

With recent speech products you can speak at a normal speaking speed (about 120 words per minute).

How does Speech Recognition work?

Speech Recognition programs use probability and mathematical functions to determine the most likely word or phrase. To identify what is spoken, the computer goes through these five steps:

• Speech input (User)
• Prefiltering
• Feature extraction
• Comparison and matching
• Text output

After you install the program, the user should undergo training but if you use Dragon NaturallySpeaking you can still achieve a reasonable result without it. Having said that, it is certainly better to go through the training process. This involves reading aloud from a prepared text for a minimum of five minutes. The computer listens to this dictation and the software analyzes how the user pronounces each word. It then stores the data and prepares a unique user profile for the user.

Further improvement in speech recognition occurs as you continue to use the software and correct recognition errors. The Speech Recogniton software becomes more accurate as it learns your particular speaking style. Dragon NaturallySpeaking offers to analyse your email and documents to assess and register the vocabulary you use. This is an additional bonus not found in Windows 7 Speech Recognition program.

Who uses Speech Recognition?

Users of speech recognition products are mostly professional people who use words extensively as part of their job. Most of them are already comfortable speaking their thoughts.  So you will find lawyers, doctors, and teachers are the early adopters.

People with disabilities limiting their use of the computer keyboard or mouse can find speech recognition programs a boon as they can perform many computer tasks ranging from writing documents to sending email. This can be liberating for them.

What are the advantages in using Speech Recognition?

1. You can “command and control” the computer simply by using your voice. This means you can command and  control:

• Menus and dialog boxes
• Edit and format text
• Launch applications
• Open and save files
• Cut and paste

Other Benefits:

1. You can also speak faster than you can type.
2. While your fingers may develop muscle fatigue, your voice will not.
3. The computer knows how to spell many of the words you speak probably better than you do.

For Speech Recognition to be effective you need a fast and powerful personal computer, up-to-date software compatable with Windows 7 and to enrol in its training program so that it can recognize your voice.

Hardware Requirements for Speech Recognition

Here is a list of essential hardware requirements to run speech recognition software efficiently.

• CPU:   Minimum 2.4 GHz Pentium® III/IV Processor IBM-compatible PC – the more powerful Dual Core and Quad Core CPUs are better.
• Operating System:     For versions higher than Version 9.5 Dragon NaturallySpeaking you will need Vista or Windows 7.
• Sound Card: Industry standard 16 bit sound card or built-in audio systems on desktops and portables, including the Creative Labs SoundBlaster® 16 or higher
• Hard Disk Requirements:     40 GIG or higher – you need 1 GIG for program storage and 40-60+ Mbytes for user voice file.
• Memory Requirements:        Minimum 1 GIG RAM – the more the better! I have 3 GIG on the laptop and 4 GIG on the Desktop.
• Microphone Headset:            A good quality USB from these sources:
  • Andrea Electronics
  • Plantronics
  • Sennheiser
  • Logitech

Speech Recognition Software

The Main Players

There have been many take-overs in the Speech Recognition industry. ScanSoft, once a big player, was taken over by Nuance. Similarly, IBM’s speech recognition software was sold off to ScanSoft in 2003. MacSpeech is powered by Dragon NaturallySpeaking engine.  This makes Nuance the lead player in Speech Recognition.

• Dragon NaturallySpeaking (Version: Standard. Preferred, Professional, Medcial and Legal)
• Microsoft Windows Speech Recognition (Windows Vista and Windows 7)
• MacSpeech (Versions MacSpeech Dictate, MacSpeech Medical and MacSpeech Legal . All powered by Dragon NaturallySpeaking engine.

Catering for Different Languages and Different English Accents

• Microsoft Speech Recognition is available in English (U.S.), English (U.K.), German (Germany), French (France), Spanish (Spain), Japanese, Chinese (Traditional), and Chinese (Simplified).
• Dragon NaturallySpeaking is available in English, French, German, Spanish, Italian and Dutch language versions. The English versions cater for the following accents, US, UK, Australian, Indian, and SE Asian.
• MacSpeech caters for 13 English language variations, nine with U.S. spelling and four with U.K. spelling.

In the next post on Speech Recognition we’ll look at these aspects:

In this post we’ll be looking at Windows 7 Speech Recognition program and considering these aspects:

• Accessing Windows 7 Speech Recognition
• Training the Software
• Reviewing Speech Recognition Commands
• Taking the Speech Tutorial