Estimating Bone Conduction Hearing Perception Using Three-Dimensional Vibration in a Head Simulator
Typ
Examensarbete för masterexamen
Master's Thesis
Master's Thesis
Program
Biomedical engineering (MPBME), MSc
Publicerad
2023
Författare
Husmark, Anton
Håkansson, Erik
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
Abstract
Assessing the performance of bone conduction (BC) hearing aids is crucial during development. At Cochlear Bone Anchored Solutions AB, one prominent method of accomplishing this is by listening to the hearing aids via head simulators, which are artificial heads equipped with accelerometers as a representation of the cochleae. A new adaptation of a head simulator that uses three-axis accelerometers and has anatomy based on a magnetic resonance imaging scan of a living subject has been developed. There are several indications in the literature that suggest measuring three-dimensional vibration of the cochlea correlates better with BC hearing than one-dimensional vibration, thus suggesting an improvement when having three-axis accelerometers since this enables the usage of all the spatial components. Therefore, this project aims to investigate how to combine three-axis vibration data into a realistic sound representation and how this can be implemented in real-time listening of BC hearing aids on the head simulator. A proposed combination model is determined through system identification with the vibration data from the head simulator and BC hearing thresholds from the subject at different positions as inputs. The correlated output is deemed to be the corresponding air conduction hearing threshold to correlate three-dimensional vibration to the experienced hearing. The combination model is implemented through finite impulse response (FIR) filters obtained from the system identification, which is analyzed in an offline environment and compared with existing algorithms. Due to the characteristics of the model, it is implementable in digital signal processing (DSP) hardware which means audio signals can be processed in real time where each component from the accelerometer is filtered and then added together. The FIR filters show a possible way of combining the three spatial components. Simulation of the signal processing of an audio file shows reasonable sound quality. It also indicates a closer connection between vibration and hearing thresholds for signals containing information from all three spatial components, compared to studying one axis at a time. Simulating signals through DSP hardware with the implemented combination filters in real time also shows non-noticeable latency. These findings indicate that the proposed model accomplishes the aim of combining three-dimensional data into a listenable audio signal. The model is however sensitive to both how the data is acquired, such as stimulation positions used for the hearing thresholds, and the quality of the data in general, such as how good of a representation the head simulator is of the subject’s real head. It is also difficult to objectively assess how realistic the sound quality is. However, literature and simulations indicate that the proposed combination model might produce a more realistic representation of BC hearing in a real-time scenario compared to only using one spatial component.