基于短时快速傅里叶变换或频谱特征提取的音频处理策略对人工耳蜗使用者性能的比较
本研究采用双盲设计,比较6名Oticon Medical Digisonic SP人工耳蜗使用者使用基于短时快速傅里叶变换的Crystalis策略与基于频谱特征提取(SFE)的实验策略在安静下单词识别、噪声下言语接收阈值、辅音区分、聆听努力、旋律轮廓识别及主观音质等方面的表现。结果显示两种策略的CI使用者性能相似,Crystalis在部分指标上更优,SFE策略可能需要更长的适应时间才能展现全部潜力。
要点
- 双盲设计比较6名Oticon Medical Digisonic SP人工耳蜗使用者使用Crystalis与SFE策略的性能。
- Crystalis在辅音区分和主观音质上优于SFE,单词识别和噪声下SRT在最后测试日两组相当。
- 通道交互较高的使用者中两种策略的性能差异更大。
核心结论
- CI使用者使用SFE和Crystalis策略的性能相似,SFE策略可能需要更长的适应时间才能展现全部潜力。
研究设计与样本
- 双盲设计,纳入6名Oticon Medical Digisonic SP人工耳蜗使用者,比较SFE与Crystalis策略。
- 单词识别和噪声下言语接收阈值在首次和最后测试日(间隔4至5天)测量,其他测试在首末测试日之间进行一次。
主要发现
- 首次测试日两种策略单词识别相似,最后测试日两种策略均改善,Crystalis改善显著更多。
- 首次测试日SFE的噪声下言语接收阈值差于Crystalis,最后测试日两者相当。
- Crystalis辅音区分得分高于SFE,SFE主观音质得分低于Crystalis。
- 两种策略的旋律轮廓识别得分和聆听努力无实质差异。
- 通道交互较高的CI使用者中,SFE与Crystalis的性能差异更大。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
OBJECTIVES: We compared sound quality and performance for a conventional cochlear-implant (CI) audio processing strategy based on short-time fast-Fourier transform (Crystalis) and an experimental strategy based on spectral feature extraction (SFE). In the latter, the more salient spectral features (acoustic events) were extracted and mapped into the CI stimulation electrodes. We hypothesized that (1) SFE would be superior to Crystalis because it can encode acoustic spectral features without the constraints imposed by the short-time fast-Fourier transform bin width, and (2) the potential benefit of SFE would be greater for CI users who have less neural cross-channel interactions. DESIGN: To examine the first hypothesis, 6 users of Oticon Medical Digisonic SP CIs were tested in a double-blind design with the SFE and Crystalis strategies on various aspects: word recognition in quiet, speech-in-noise reception threshold (SRT), consonant discrimination in quiet, listening effort, melody contour identification (MCI), and subjective sound quality. Word recognition and SRTs were measured on the first and last day of testing (4 to 5 days apart) to assess potential learning and/or acclimatization effects. Other tests were run once between the first and last testing day. Listening effort was assessed by measuring pupil dilation. MCI involved identifying a five-tone contour among five possible contours. Sound quality was assessed subjectively using the multiple stimulus with hidden reference and anchor (MUSHRA) paradigm for sentences, music, and ambient sounds. To examine the second hypothesis, cross-channel interaction was assessed behaviorally using forward masking. RESULTS: Word recognition was similar for the two strategies on the first day of testing and improved for both strategies on the last day of testing, with Crystalis improving significantly more. SRTs were worse with SFE than Crystalis on the first day of testing but became comparable on the last day of testing. Consonant discrimination scores were higher for Crystalis than for the SFE strategy. MCI scores and listening effort were not substantially different across strategies. Subjective sound quality scores were lower for the SFE than for the Crystalis strategy. The difference in performance with SFE and Crystalis was greater for CI users with higher channel interaction. CONCLUSIONS: CI-user performance was similar with the SFE and Crystalis strategies. Longer acclimatization times may be required to reveal the full potential of the SFE strategy.
引用
- Zhang Y, Johannesen PT, Molaee-Ardekani B, et al. Comparison of Performance for Cochlear-Implant Listeners Using Audio Processing Strategies Based on Short-Time Fast Fourier Transform or Spectral Feature Extraction. Ear Hear. 2024. PMID: 39680489. doi:10.1097/AUD.0000000000001565
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。