A Novel Unsupervised Spectral Clustering for Pure-Tone Audiograms towards Hearing Aid Filter Bank Design and Initial Configurations

Abstract

The current practice of adjusting hearing aids (HA) is tiring and time-consuming for both patients and audiologists. Of hearing-impaired people, 40–50% are not satisfied with their HAs. In addition, good designs of HAs are often avoided since the process of fitting them is exhausting. To improve the fitting process, a machine learning (ML) unsupervised approach is proposed to cluster the pure-tone audiograms (PTA). This work applies the spectral clustering (SP) approach to group audiograms according to their similarity in shape. Different SP approaches are tested for best results and these approaches were evaluated by Silhouette, Calinski-Harabasz, and Davies-Bouldin criteria values. Kutools for Excel add-in is used to generate audiograms’ population, annotated using the results from SP, and different criteria values are used to evaluate population clusters. Finally, these clusters are mapped to a standard set of audiograms used in HA characterization. The results indicated that grouping the data in 8 groups or 10 results in ones with high evaluation criteria. The evaluation for population audiograms clusters shows good performance, as it resulted in a Silhouette coefficient >0.5. This work introduces a new concept to classify audiograms using an ML algorithm according to the audiograms’ similarity in shape.

Data-driven audiogram classifier using data normalization and multi-stage feature selection

“…The authors as discussed in the previous paper 15 decided to remove the set of audiograms that represent normal hearing levels. These levels are removed since the algorithm is developed to assist in configuring the hearing aids for patients who are experiencing hearing loss.…”

“…Then, ML unsupervised spectral clustering was applied to classify these audiograms into classes according to similarity in the shape 15 . The authors clustered the quantized data into 7–11 clusters, and based on the statistical test conducted, 10 clusters are selected as the optimum number of clusters as this number of clusters gives the highest criteria values.…”

“…The data’s nature is of high overlapping which means the different shapes cannot be captured by a low number of classes. A detailed description to the authors’ previous work on unsupervised spectral clustering algorithm, how it was implemented and its performance evaluation can be found in 15 .…”

“…The authors chose to use different ranges to clearly show the difference in the detected shapes.

Figure 2 Audiograms clusters based on spectral clustering 15 .

…”

“…In the current study, a supervised machine learning technique is developed based on the annotated data from the unsupervised spectral clustering done by 15 . This data source contains samples of clean data with no outliers, no redundant values, no missed values and it is in the format that is suitable for use.…”