Word Recognition in Continuous Speech and Speaker Independent by Means of Recurrent Self-Organizing Spiking Neurons

Artificial neural networks have been applied successfully in many static systems but present someweaknesses if patterns involve a temporal component. Let’s note for example in speech recognition orcontextual information, where different of the time interval, is crucial for comprehension. Speech,being a temporal form of sensory input, is a natural candidate for investigating temporal coding inneural networks. It is only through comprehension of the temporal relationship between differentsounds which make up a spoken word or sentence that speech becomes intelligible. In fact we presentin this paper presents three variants of self-organizing maps (SOM), the Leaky Integrators Neurons(LIN), the Spiking_SOM (SSOM) and the recurrent Spiking_SOM (RSSOM) models. The proposedvariants is like the basic SOM, however it represents the characteristic to modify the learning functionand the choice of the best matching unit (BMU). The case study of the proposed SOM variants is wordrecognition in continuous speech and speaker independent. The proposed SOM variants show goodrobustness and high word recognition rates.