Mobile edge computing based environment for active noise control on smartphone

Abstract

Violation of privacy while talking on the phone is a crucial problem especially when there is a need to communicate personal and confidential information during the presence of others. Sound masking, which is a method based on adding disturbing sound to the speech, is an effective solution when it comes to speech intelligibility. However, sound masking introduces adverse impacts to both far-end and near-end users. The generated sound masking creates a Feedback Effect for the far-end user and Lombard Effect for the near-end user

thus, causing degradation of communication quality for both the far-end and near-end users. Those effects can be eliminated through Active Noise Control. However, the current smartphones are unable to endure the high computation power of ANC algorithms. Therefore, a new approach for ANC algorithm that separates the high computation power component and offloads it to a low-latency server is proposed in this research the chosen ANC algorithm can be divided into two processes: convolution process and weighting process. The convolution process between reference signals and weighting filters, which requires less computation power, is performed in the smartphone CPU, and the weighting process for weighting filter derivation, which requires more computation power, is performed in the server. In this research, Amazon Web Service which provides a platform for MEC (Mobile Edge Computing) for computation offloading applications through its service called AWS Wavelength is used in this research.