Using dynamic synapse based neural networks with wavelet preprocessing for speech applications

One major problem in the field of voice recognition is noise robustness. This project involved the design of a system that is both robust in the presence of noise as well as being capable of two major tasks in voice recognition: speaker verification on a closed set of speakers and speech recognition on a closed set of speakers using a set of command phrases. The system uses a wavelet processing technique that allows for either speaker-dependent or word-dependent feature set extraction. Both tasks are accomplished using a dynamic synapse based neural network with noise resistance properties that is trained using a genetic algorithm technique. Using these techniques, the system was able to perform the speaker verification task as well as the speech recognition task without being adversely affected by normal levels of noise, and perform verification despite low variability between speakers or words.