Literature Review on State-of-the-Art Algorithms on Low Power Hearing-Aid Related Audio Processing

Abstract

In today world, hearing-aid devices are crucial for people with defective hearing ability. In the heart of such device lies the digital signal processor (DSP), which takes responsibility for processing the incoming speech and producing the undistorted audio signal to the user. The hearing-aid devices must be small enough to be worn in human ears, yet powerful to compute a large amount of data with low latency and high accuracy. Last but not least, hearing-aid batteries should last for several days to meet the customer satisfaction. Therefore, they should be optimized in terms of performance and energy consumption.

In this thesis, the state-of-the-art audio processing algorithms that are widely utilized in digital hearing-aid devices are identified. The potential algorithms are divided into different parts according to the DSP functionality. For each part, the fundamentals of the most applicable methods and some of their revised versions are explained in detail.

Based on the selected domains and algorithms, this thesis work explores some efficient low power implementations in real digital hearing aids. The collection is again separated into subsections corresponding to the DSP processing segments. Subsequently, the cutting-edge low-power implementations of the whole DSP for hearing-aid purposes are classified based on the hardware architecture.

In summary, most of the findings belong to feedback cancellation and noise reduction categories because those sections determine the quality of the audio output. In addition to the signal characteristics, a high output signal amplitude is desired, which is why beamforming algorithms play an important role in hearing-aid processors. As regards the technology, Application-Specific Instruction-set Processor (ASIP) with numerous hardware accelerators dominates the recent low-power hearing-aid implementations because of its performance and power efficiency.