Chapter 7 — Hearing aids

DSP, WDRC, directional processing, feedback cancellation.

A modern hearing aid is a small but extraordinarily sophisticated computer attached to the ear. Two omnidirectional microphones sample the acoustic environment 20,000–48,000 times per second. The samples flow through a digital signal-processing pipeline that performs spectral analysis, frequency-dependent gain shaping, dynamic-range compression, directional beamforming, noise reduction, feedback cancellation, and frequency lowering — all at audio latencies under 6 milliseconds, all in a few cubic centimetres of plastic, all running on a battery that lasts several days. The DSP chip in a 2026 hearing aid runs at 10s of GFLOPS and consumes about 1 mW.

This chapter develops the modern hearing aid algorithmically. We start with the DSP pipeline as a whole — the block diagram from microphone to receiver — then drill into three of its most important components: wide-dynamic-range compression (the technique that fits the 80 dB range of real-world acoustic levels into the patient’s narrowed residual dynamic range), directional microphones (the two-mic beamformer that gives 3–6 dB of SNR improvement against background noise), and feedback cancellation (the adaptive-filter algorithm that lets a small open-fit hearing aid apply 40 dB of gain without howling).

The chapter is the longest in this book; hearing aids are the workhorse intervention that the rest of the toolkit ultimately exists to support. Four lessons: