A time delay neural network for recognition the stop and nasal Arabic speech

الملخص EN

In this work.

A Time Delay Neural Network for Recognition the Stop and Nasal Arabic Sounds is presented.

Processing for Arabic Speech in time domain, four parameters explain to each words spoken bv(16) speakers(8 male and 8 female) and given to network as the input in training stage.These parameters are Zero Crossing, Energy.

Average Magnitude and max Fundamental Frequency.

Hamming Window used for this work and recording with the sound blaster adapter sample represented as 8 bit integer which frequency(digitized rate) was(8000) HZ.

The training stage in the network take along time(three weeks) to recognized the Unvoiced stop Arabic speech ) and Voiced stop Arabic speech (m , e) and a short time to recognized the Nasal Arabic speech (u ).The neural network recognized rate for Arabic phoneme is 96% after training.