A survey of transfer functions for independent speaker vowel recognition using neural networks
Abstract
Artificial Neural Networks (ANNs) were born in a desire to simulate the human brain and solve problems that computers can't solve. The ability of them to acquire store and use knowledge obtained by experience give them the ability to learn and adapt by example, taking into account previous errors. This characteristic makes Neural Networks a very important tool in pattern recognition. The advances in pattern recognition and in specific in voice recognition have been very limited since the difference in pronunciation because of many different factors including age, gender, health, weather, and others can affect the performance of those systems. This thesis will implement some activation functions and the results obtained by using different functions will be observed and compared to verify any improvement with the performance of an Independent Speaker Vowel Recognition Neural Network.
Subject Area
Computer science|Artificial intelligence
Recommended Citation
Calles Cowell, Erubey, "A survey of transfer functions for independent speaker vowel recognition using neural networks" (2004). ETD Collection for University of Texas, El Paso. AAIEP10525.
https://scholarworks.utep.edu/dissertations/AAIEP10525