29 lines
6.5 KiB
TeX
29 lines
6.5 KiB
TeX
\section{Introduction}
|
|
\subsection{Motivation}
|
|
According to the World Health Organization (WHO), around 1.6 billion people over 14 years worldwide suffer from any kind of hearing loss. Included in this 1.6 billion people, around 430 million suffer from disabling hearing loss (up to deafness), requiring rehabilitation. In the case of disabling hearing loss, the possibility of using an implant system solution has revolutionized auditory rehabilitation by restoring partial hearing. Despite steady progress in implant technology over the past decades, the system still faces its limitations. Complex auditory environments, like static noises overlain by a person speaking, can still propose a considerable challenge for users of auditory implants compared to people with a healthy hearing. \\ \\
|
|
Therefore, the improvement of implant performance in regard to the suppression of disturbance noises is therefore a crucial step in the development of more user-friendly implant solutions which provide users with more natural sound perception and greater listening comfort.
|
|
\\ \\
|
|
By addressing these challenges, this work aims to contribute to the next generation of cochlear implant technology, ultimately enhancing the auditory experience and quality of life for people with severe hearing impairments.
|
|
\subsection{Introduction to cochlear implant systems}
|
|
A cochlear implant (CI) System is a specialized form of hearing aid, used to restore partly or complete deafness. In contrary to standard hearing aids, CI's do not just amplify the audio signal received by the ear, but stimulate the auditory nerve itself directly through electric pulses.\\ \\
|
|
Usually, a CI system consists out of an external processor with a microphone (``audio processor'') receiving the ambient audio signal, processing it, and then transmitting it inductively via a transmission coil through the skin to the cochlear implant itself, implanted on the patient's skull (see Figure \ref{fig:fig_synchrony}). The CI stimulates the auditory nerves inside the cochlear through charge pulses, thus enabling the patient to hear the received audio signal as sound.\\
|
|
\begin{figure}[H]
|
|
\centering
|
|
\includegraphics[width=0.6\linewidth]{Bilder/fig_synchrony.png}
|
|
\caption{Sketch of a MED-EL Synchrony Cochlear Implant with a Sonnet 3 Audio Processor \cite{source_synchrony}}
|
|
\label{fig:fig_synchrony}
|
|
\end{figure}
|
|
\noindent The pulse transmission to the cochlear is realized through a silicone electrode with embedded metal contacts. Said electrode is inserted into the cochlear through a drilled hole in the bone, where, depending on the insertion depth, different contact areas stimulate different parts of the frequency spectrum of the hearing sense. The smaller end of the electrode array inserted deep into the cochlear stimulates low frequencies, whereas the larger part at the beginning of the array stimulates high frequencies. (see Figure \ref{fig:fig_electrode}).
|
|
\begin{figure}[H]
|
|
\centering
|
|
\includegraphics[width=0.8\linewidth]{Bilder/fig_electrode.jpg}
|
|
\caption{Visualization of a MED-EL electrode inserted into a human cochlear. \cite{source_electrode}}
|
|
\label{fig:fig_electrode}
|
|
\end{figure}
|
|
\noindent As for any head worn hearing aid, the audio processor of a CI system does not only pick up the desired ambient audio signal, but also any sort of interference noises from different sources. This circumstance leads to a decrease in the quality of the final audio signal for the user. Reducing this interference noise through adaptive noise reduction, implemented on a low-power digital signal processor, which can be powered within the electrical limitations of a CI system, is the topic of this master's thesis.
|
|
\subsection{Implementation of Adaptive Noise Reduction in Cochlear Implant Systems}
|
|
The above problem description of noise interference shows the need of further improvement of CI systems in this regard. For persons with a healthy hearing sense, the addition of noise to an observed signal may just mean a decrease in hearing comfort, whereas for aurally impaired people it can make the difference in the basic understanding of information. As everyday environments present fluctuating background noise - from static crowd chatter to sudden sounds of different characteristics — that can severely degrade speech perception, the ability to suppress noise is a crucial benefit for users of cochlear implant systems. \\ \\
|
|
Adaptive noise reduction (ANR) (also commonly referred as adaptive noise cancellation (ANC)), is an advanced signal processing technique that adjusts the parameters of a digital filter to suppress unwanted noise from a signal while preserving the desired target signal. In contrary to static filters (like a high- or low-pass filter), ANR uses real-time feedback to adjust said digital filter to adapt to the current circumstances.\\ \\
|
|
The challenge in the implementation of ANR in CI systems lies in the limited capacities. As the CI system is powered by a small battery located in the audio processor, energy efficiency is crucial for a possible solution of the described problem of noise interference. Any approach to a reduction of interference noise must be highly optimized with regard to computing power and implemented on dedicated low-power hardware, being able to be powered within the limitations of a CI system.\\ \\
|
|
The main solution concept of this thesis is the optimization of the adaptive filter of the ANR algorithm in combination with the used low-power hardware. Its goal is, to deliver the best possible result in interference noise reduction while still being able to be powered by the limited resources of a CI system. Different variants, like the fully adaptive filter, the hybrid static/adaptive filter and different optimization approaches of the latter one are low-level simulated on the dedicated digital signal processor. Especially, the different optimization strategies of the hybrid static/adaptive filter algorithm shall be evaluated and compared in regard of their required computing power, and therefore, their required power consumption. Depending on the kind of interference noise, the frequency and the intensity, a promising optimization approach is the reduction of adaptation steps per sample while still maintaining an adequate quality of the filtered audio signal.\\ \\
|
|
Due to the fact, that the CI system is powered by a battery with a relatively small capacity, the firmware is required to work with the least power possible. Therefore, optimization in regard to a minimization of needed processor clocks is aimed for. |