Potential Problems
| Home | Title Page | Introduction | Background Information | Potential Benefits | Potnetial Problems | Conclusion | Bibliography | IT 103 Research Paper PDF | Research Paper Blog | Research Paper Wiki |
All new advancements come with their flaws and in the case of automatic speech recognition, there is quite a few. First of all, everyone’s voices are different that is what makes us unique but it also generates problems. “The fundamental frequencies of speech sounds uttered by woman are about an octave higher than those of men, and those of children are even higher” (Ainsworth, 1939). Not only does voice differ person to person but it also depends on our age and our sex. When a toddler, they are just learning to speak so they do not pronounce sounds and words properly, and then children go through puberty where if a boy, their voice will squeak and change dramatically over a short period of time until settled into their so called set voice for life. How is voice recognition suppose to understand all these different stages of growth all at one time on top of accents and mumbling, it cannot. Also, with so much migration there are so many different accents;
"People from different parts of the country and different social an economic phenome from another in certain words, and changes in the rhythm and inonation of the utterance. This variability is even greater with people speaking a second language. The competence with which it is spoken depends on the motivation, intelligence, and perceptual and motor skills of the speaker, and also of the age at which the second language was learned (Ainsworth, 1939)."
With the example that Bostonians do not pronounce their “R’s” and that New Yorker’s replace their “E’s” for “A’s,” its merely impossible for technology to comprehend accents of Americans on top of other countries if real live people have a hard time understanding each other. In addition, white noise is an issue, which is “background noise as the caller interacts with the system,” that may give false triggers to the recognition or the software that may not recognize at all what the speaker is saying (Schelps). There are still many flaws to be adjusted to the program to make it a more credible software.