Challenges in Voice Recognition Technology
By admin
Voice recognition systems actually cannot be saying 100 percent working. Several factors can reduce accuracy, for example in terms of signal to noise ratio is low. While programs need to hear words were spoken clearly and a little extra sound coming into the program. Its use also requires a space that is not crowded with people and quiet, plus quality microphone and the mouth is positioned as close as possible to the funnel. Soundcard use that function to process the input from the microphone should not be too low quality. When low-quality soundcard has no protection from electrical signals produced by other computer components.
As a result, sometimes often cause hum or hiss in the signal that would be processed. The system is so far also having difficulty in separating the sound from many users. If the technology would be used in a conversation or meeting, the system was not able to translate the sound to the max. ?Even the fastest personal computer sometimes still having trouble with complicated phrases and commands so that the slow response time significantly. List of words that need this program helped take a big enough hard drive space. But it can be resolved if you see a computer disk space technology development is very rapid. The last factor is the weakness in recognizing homonym.
Homonym is two words that are spelled differently and have a meaning that is not the same but came out the same. This makes the voice recognition system it is impossible to distinguish the two words based on sound alone. However, training routine on the system to improve its accuracy. ?For several decades, computer scientists difficulty in developing methods pioneered systems that Alexander Graham Bell was inspired desire to transmit the word to his wife who was deaf. All of that is due to technological limitations. Since 1990, this technology is utilized The Defense Advanced Research Projects Agency (DARPA), agent of the United States military technology development. The institution that has three teams of researchers in the Global Autonomous Language Exploitation (Gale) that retrieve information from foreign broadcasts and translates. DARPA is also supporting the development of TRANSTAC, a program that makes a person, especially soldiers, to communicate with local people who cannot speak English. Gale as a universal translator was sometimes difficult to apply because of constrained slang, dialect, accent, and background noises. For example, Arabic sometimes uses only one word to express an intention, whereas in other languages may form a sentence. But voice recognition system can be more accurate it is possible. Because the statistical model to let the computer decide what kind of person who is talking and capture the implied meaning behind the word. The computer experts also recognize the voice recognition system as the progress of the most progressive and capable of replacing hardware like keyboard that makes the hands sore and tiring.