Influence of microphone soiling on syllable recognition in cochlear implants: simulation and recognition in noise

Abstract

Background: The microphone is the first link of the sound processing chain and if there is dirt on the membrane or in the input port, the frequency response is modified.

Objective: The effect of microphone cleaning in cochlear implants (CI) was studied by measuring speech perception scores in the noise of CI users and of normal hearing (NH) subjects with a CI simulator.

Material and methods: Sixty-one adult CI users and 20 NH were tested. In CI subjects, speech comprehension scores in noise were compared before and after cleaning microphone ports with one of two different cleaning procedures. NH listeners were tested using a vocoder that simulated soiled microphones. Two main coding schemes were considered, channel-picking (CP) and fixed-channel (FC).

Results: With NH subjects, the effect of the coding strategy and degree of soiling was statistically significant (p < 10⁻⁴); for the intermediate signal to noise ratios (SNRs), the recognition percentages were clearly affected. With CI users, cleaning the microphones significantly improved the recognition scores (p < 10⁻⁴). The two cleaning procedures considered turned out to be significantly different (p < 10⁻⁴).

Conclusions: It is suggested to perform a regular microphone port cleaning to avoid deep soiling which leads to a decrease in speech understanding.

Chinese abstract

背景：麦克风是声音处理链的第一个环节。如果薄膜或输入端口有污垢, 频率响应就会改变。

目的：测量人工耳蜗（CI）使用者和使用CI模拟器的正常听力（NH）受试者的语音感知评分, 研究人工耳蜗（CI）内麦克风清洁的效果。

材料与方法：对六十一名成年CI使用者和20名NH进行测试。在CI受试者中, 比较使用两种不同清洁方法清洁麦克风输入端口前后的噪声语音理解评分。对使用声码器模拟脏麦克风的NH受试者进行了测试。主要考虑两种编码方案：信道选择（CP）和固定信道（FC）。

结果：编码方案和污染程度对NH受试者的识别率有显著影响（p＜10⁻⁴）, 对中等信噪比（SNR）的识别率有显著影响。对于CI使用者, 清洁麦克风显著提高识别评分（p＜1 0⁻⁴）。所考虑的两种清洁程序结果显著不同（p ＜1 0⁻⁴）。

结论：建议定期进行麦克风端口清洗, 以避免导致语音理解降低的深层污染。

Keywords:

• Cochlear implant
• microphone soiling
• cleaning procedures
• simulation
• CI users and NH subjects
• syllable recognition in noise
• coding strategies

Disclosure statement

No potential conflict of interest was reported by the authors.