A Comprehensive Review of Beamforming-Based Speech Enhancement Techniques, IoT, and Smart City Applications

Speech recognition from a distance, also known as far-field automatic speech recognition, uses machine learning for processing. However, environmental conditions often corrupt speech recorded from a distance, causing disturbances. To obtain desired speech from corrupted signals and to enhance the quality of the signal, various speech enhancement techniques are used, such as de-reverberation, source separation, denoising, and acoustic beamforming. The aim is to design a robust and multi-condition adaptive system in far-field-based automatic speech recognition systems. This review paper focuses on speech enhancement for the future of speech with progressive technologies like deep learning and machine learning. It highlights the extensive research on beamforming-based speech enhancement over the past few years, based on different techniques, performance, advantages, limitations, and scope for improvement. Finally, this paper explores the smart city applications that benefited from speech enhancement and beamforming.