Quantum Machine Learning and Spectrogram Fusion for Speech Recognition

This study evaluates the use of machine learning and quantum spectrogram image fusion in speech recognition. More specifically, a quantum simulation circuit is designed to process and fuse Fourier transform and linear prediction generated spectrograms. In addition, deep neural networks are used to obtain speech classification results. The performance of the quantum computing simulation model is evaluated and compared to the classical speech recognition method in terms of classification accuracy. Moreover, we assess the merit of quantum spectrogram fusion for speech recognition tasks in terms of confusion matrices. We report classification results with the Google Speech Commands database and we briefly examine the effects of quantum noise.