Real Time Speech Processing Algorithms For Smartphone Based Hearing Aid Applications PDF Download

Hearing impairment is a pervasive problem which occurs due to the detrimental damage caused to the inner ear. Assistive Hearing Devices such as Cochlear Implants (CIs) and Hearing Aids (HAs) are designed to restore hearing, personalize rehabilitation, and enrich the listening experience. Although signal processing and machine learning research has greatly improved audio processing, the rigid design requirements of commercial CI sound processors make it difficult to explore novel algorithms for research investigations and conduct longitudinal studies. This thesis presents the design, development, clinical evaluation, and applications of CCi-MOBILE, a computationally powerful signal processing testing platform built specifically for researchers in the CI/HA field along with implementing multiple additional features to the platform. This custom-made, portable research platform allows researchers to design and perform complex speech processing algorithm assessment offline and in real-time through user-friendly, software-mediated open-source tools with implants manufactured by Cochlear Corporation. The design includes a lightweight custom circuit board comprising of an on-board FPGA to be used in conjunction with a computing platform such as a PC/tablet/laptop/smartphone based on the requirement of CI/HA signal processing algorithms. The processing pipeline for CI and HA stimulation is discussed followed by results from an acute study with implant users’ speech intelligibility in quiet and noisy conditions. The platform supports testing of algorithms for unilateral, bilateral, and bimodal hearing impairment. A major obstruction to accurate source localization for bimodal and bilateral CI users is the distortion of interaural time and level difference cues (ITD and ILD), and limited ITD sensitivity. Various CI research interfaces developed by either academic or industry sponsored research teams support proposed signal processing and psychoacoustic investigations but have limited ability to efficiently validate bimodal and/or bilateral algorithms. To overcome such challenges; verification, and validation of the synchronized bilateral (electric-electric) and bimodal (electric-acoustic) outputs is performed, in an authenticated and efficient way, to support localization algorithmic and experimental investigations. It has been hypothesized that variable stimulation rate for exciting the electrode array can aid for better speech perception and increased spectral information. Hence, a new multi-rate implant strategy including time-varying stimulation rates has been proposed in this work. Lastly, expanding the capabilities of the platform to ensure long-term sustainability, a real-time data streaming link between the platform and a cloud-based data repository is established to enable remote-test facilities along with an algorithm implementation and testing in naturalistic environments. We discuss implementation feasibility, and hypothesized performance of these approaches individually, and collectively, on the perceptual benefit for researchers working towards the welfare of the hearing-impaired community.