Improved performance with automatic sound management 3 in the MED-EL SONNET 2 cochlear implant audio processor

Abstract

Objectives: The SONNET 2 audio processor features ambient noise reduction (ANR), transient-noise reduction (TNR), and adaptive intelligence (AI). The primary aim of this study was to evaluate if using these features improves speech perception in noise, subjective listening effort, and sound quality.

Design: In this prospective longitudinal study, twenty adult SONNET users were fitted with the SONNET 2 audio processor, configured either as a default SONNET (no ANR/TNR/AI), with mild ANR/TNR, with strong ANR/TNR, with mild AI, and with strong AI. Speech perception in noise was assessed in speech and stationary noise from the front (S0N0); speech, stationary noise, and transient noise from the front (S0N0T0); and speech from the front in spatially-distributed stationary noise (S0N±45N±135). Listening effort, subjective sound quality, and device/setup preference were assessed.

Results: In the S0N0 setup, speech perception in noise was significantly better with the SONNET 2 when using ANR/TNR in the mild setup than with the SONNET or the SONNET 2 in the default SONNET configuration. In the S0N±45N±135 setup, speech understanding was significantly better in all four SONNET 2 configurations than with the SONNET or the SONNET 2 in the default SONNET configuration (a 1.26-2.55 dB SRT80 benefit). Subjects tolerated consistently lower signal-to-noise values with the SONNET 2 configurations using ANR/TNR than with the default SONNET configuration in all listening effort categories. All SONNET 2 configurations using ANR/TNR were preferred and better rated in speech in stationary and/or transient noise compared to the default SONNET configuration. Sound quality and pleasantness were better in those SONNET 2 configurations. Subjects strongly preferred the SONNET 2 configurations over the SONNET configuration.

Conclusions: The new front-end features implemented in the SONNET 2 audio processor objectively improve speech perception in noise. Subjects preferred the SONNET 2, over the SONNET, in the presence of stationary and transient noise.

Fig 1

Speech test setups for (a) speech or music testing in quiet (S0 or M0), (b) speech in noise S0N0, (c) speech in noise S0N0T0, (d) speech in noise S0N±45N±135, and (e) speech in noise S0N±45N±135T135 (note: transient noise presented ipsilaterally to ear with the CI). Orange speaker icons indicate target signal (speech or music), black speaker icons indicate stationary noise, and the blue speaker icons indicate transient noise.

Fig 2. Speech intelligibility in quiet with each configuration.

Boxplots show median values as well as first and third quartiles. Whisker bars show minimum and maximum values within 1.5 interquartile ranges below and above the first and third quartile. Diamonds show outliers.

Fig 3. SRT80 observed in the OLSA S0N0 for each configuration.

A cross marks a significant difference compared to S2.Sonnet. A star marks a significant difference compared to the SONNET. Boxplots show median values as well as first and third quartiles. Whisker bars show minimum and maximum values within 1.5 interquartile ranges below and above the first and third quartile. Diamonds show outliers.

Fig 4. SRT80 observed in the OLSA S0N0T0 for each test configuration.

Boxplots show median values as well as first and third quartiles. Whisker bars show minimum and maximum values within 1.5 interquartile ranges below and above the first and third quartile. Diamonds show outliers.

Fig 5. SRT80 observed in the OLSA S0N±45N±135 for each configuration.

A cross marks a significant difference compared to S2.Sonnet. A star marks a significant difference compared to the SONNET. Boxplots show median values as well as first and third quartiles. Whisker bars show minimum and maximum values within 1.5 interquartile ranges below and above the first and third quartile. Diamonds show outliers.

Fig 6. ACALES results (SNR difference, see text) for a) S2.Mild vs SONNET in S0N0, b) S2.Strong vs SONNET in S0N0, c) S2.Strong vs. S2. Mild in S0N0, d) S2.Mild vs SONNET in S0N±45N±135, e) S2.Strong vs. SONNET in S0N±45N±135, f) S2.Strong vs. S2. Mild in S0N±45N±135.

Positive scores indicate a higher tolerated noise level (i.e., better performance). Boxplots show median values as well as first and third quartiles. Whisker bars show minimum and maximum values within 1.5 interquartile ranges below and above the first and third quartile. Diamonds show outliers.

Fig 7. Product-specific questionnaire results.

Figures are the S2.Sonnet vs. 7a) S2.Mild, 7b) S2.Strong, 7c) S2.AImild, 7d) S2.AIstrong. A point indicates that this percentage of users preferred either the SONNET 2 configuration given on the abscissa, had no preference (None), found the question not applicable (NA), or did not rate (Missing) for the specific listening environment. Higher scores indicate greater preference. Diamonds indicate outliers.