More Math-Digital Signal Processing
We have focused on describing just the first step of the processing – decomposing the signal into frequency banded information. Some CIs use alternative techniques, but the FFT is certainly one of the most viable options for this step. Regardless of the algorithm used, however, it is only the first step. There is way more math to be done in the speech processor to make the CI work. The signal requires further processing and shaping. Noise filtering can help enable the user to focus on sounds of interest. But it is not just frequency of information that is important – timing, transients, and the cadence all play important roles in being able to distinguish the fine details of phonics.
Very complex algorithms have been designed that are meant to optimize performance in many different environments, generally with a goal of increasing the user’s ability to understand spoken language. As processing capability improves, the designers of cochlear implants take advantage of that processing capability to improve the quality of the listening devices. The field of digital signal processing (DSP) is critical to the continued improvement of these devices, and DSP means mathematics. To learn more details about the function of DSP and the CI speech processor, a good starting point is a 1998 review tutorial MIMICKING THE HUMAN EAR by Philip Loizou. His follow-up article from 2006 outlines the latest developments in the field:
SPEECH PROCESSING IN VOCODER-CENTRIC COCHLEAR IMPLANTS
To learn more about DSP, students might consider looking at the textbook ENGINEERING OUR DIGITAL FUTURE written to be part of a high school curriculum.