Keyword spotting in continuous speech using convolutional neural network

Abstract

Keyword spotting is a process of finding some specific words or phrases in recorded speeches by computers. Deep neural network algorithms, as a powerful engine, can handle this problem if they are trained over an appropriate dataset. To this end, the football keyword dataset (FKD), as a new keyword spotting dataset in Persian, is collected with crowdsourcing. This dataset contains nearly 31000 samples in 18 classes. The proposed continuous speech synthesis method allows the FKD dataset to be extended for practical use with applications that process continuous speeches. Besides, we proposed a lightweight architecture called EfficientNet-A0 (absolute zero) by applying the compound scaling method on EfficientNet-B0 for keyword spotting task. Finally, the proposed architecture is evaluated with various models. In our experimental results, the proposed EfficientNet-A0 and ResNet models outperform other models on the FKD dataset.