The Silicon Cochlea

Abstract

The debate about whether we hear through a mechanism involving a bank of bandpass filters or through a traveling-wave mechanism was settled by von Bekesy in favor of traveling waves. Experience gained in the engineering design and analysis of a silicon cochlea suggests why we may have evolved a traveling-wave mechanism for hearing: Distributed traveling-wave amplification is a vastly more efficient way of constructing a wide-dynamic-range frequency analyzer than is a bank of bandpass filters. Traveling-wave mechanisms are, however, more susceptible to parameter variations and noise. Collective gain control and an exponentially tapering set of filters can solve both of these problems; the biological cochlea implements both of these solutions. These engineering studies suggest that the biological cochlea, which is capable of sensing sounds over 12 orders of magnitude in intensity while dissipating only a few microwatts, is an extremely well-designed sensing instrument. We illustrate the engineering principles in the cochlea by demonstrating a 117-stage adaptive silicon cochlea that operates over six orders of magnitude in intensity over a 100 Hz-10 kHz frequency range while only consuming 0.5 mW of power. This artificial cochlea with automatic gain control and a low-noise traveling-wave amplifier architecture has the widest dynamic range of any artificial cochlea built to date.