Long Short-Term Memory Recurrent Neural Network for Automatic Speech Recognition

Automatic speech recognition (ASR) is one of the most demanding tasks in natural language processing owing to its complexity. Recently, deep learning approaches have been deployed for this task and proven outperform traditional machine such as Artificial Neural Network (ANN). In particular, deep-learning methods long short-term memory (LSTM) achieved improved ASR performance. However, method limited continuous input streams. Traditional LSTM requires four (4) linear layers (multilayer perceptron (MLP) layer) per cell with a large bandwidth each sequence time step. cannot accommodate many computational units required streams because system does not sufficient feed units. study, an enhanced recurrent neural network (RNN) model was proposed resolve shortcoming. model, RNN incorporated “forget gate” block allow resetting states at beginning sub-sequences. This enables process efficiently without necessarily increasing bandwidths. standard architecture modified effectively use parameters. Some CNN-based sequential models were used on same dataset, compared model. LSTM-RNN outperformed other accuracy 99.36% well-established public benchmark spoken English digit dataset.