Electrical Engineering Department, Najafabad Branch, Islamic Azad University, Najafabad, Iran.
Department of Biomedical Engineering, Faculty of Engineering, the University of Isfahan, Isfahan, Iran.
Comput Methods Programs Biomed. 2018 Jun;159:103-109. doi: 10.1016/j.cmpb.2018.03.003. Epub 2018 Mar 12.
Cochlear implants (CIs) are electronic devices restoring partial hearing to deaf individuals with profound hearing loss. In this paper, a new plug-in for traditional IIR filter-banks (FBs) is presented for cochlear implants based on wavelet neural networks (WNNs). Having provided such a plug-in for commercially available CIs, it is possible not only to use available hardware in the market but also to optimize their performance compared with the-state-of-the-art.
An online database of Dutch diphone perception was used in our study. The weights of the WNNs were tuned using particle swarm optimization (PSO) on a training set (speech-shaped noise (SSN) of 2 dB SNR), while its performance was assessed on a test set in terms of objective and composite measures in the hold-out validation framework. The cost function was defined based on the combination of mean square error (MSE), short‑time objective intelligibility (STOI) criteria on the training set. Variety of performance indices were used including segmental signal- to -noise ratio (SNRseg), MSE, STOI, log-likelihood ratio (LLR), weighted spectral slope (WSS), and composite measures CC and C. Meanwhile, the following CI speech processing techniques were used for comparison: traditional FBs, dual resonance nonlinear (DRNL) and simple dual path nonlinear (SPDN) models.
The average SNRseg, MSE, and LLR values for the WNN in the entire data set were 2.496 ± 2.794, 0.086 ± 0.025 and 2.323 ± 0.281, respectively. The proposed method significantly improved MSE, SNR, SNRseg, LLR, C C and C compared with the other three methods (repeated-measures analysis of variance (ANOVA); P < 0.05). The average running time of the proposed algorithm (written in Matlab R2013a) on the training and test sets for each consonant or vowel on an Intel dual-core 2.10 GHz CPU with 2GB of RAM was 9.91 ± 0.87 (s) and 0.19 ± 0.01 (s), respectively.
The proposed algorithm is accurate and precise and is thus a promising new plug-in for traditional CIs. Although the tuned algorithm is relatively fast, it is necessary to use efficient vectorized implementations for real-time CI speech signal processing.
人工耳蜗是一种电子设备，可以为患有深度听力损失的失聪个体恢复部分听力。本文提出了一种基于小波神经网络（WNN）的新型插件，用于传统的耳蜗植入物中的 II 滤波器组（FB）。通过为市售的人工耳蜗提供这样的插件，不仅可以使用市场上现有的硬件，还可以对其性能进行优化，使其达到最先进的水平。
本研究使用了荷兰语电话感知的在线数据库。WNN 的权值通过粒子群优化（PSO）在训练集（信噪比为 2dB 的语音噪声（SSN））上进行调整，同时在测试集上使用客观和综合指标在保留验证框架中评估其性能。代价函数是基于训练集上均方误差（MSE）和短时间客观可懂度（STOI）标准的组合来定义的。我们使用了多种性能指标，包括分段信噪比（SNRseg）、MSE、STOI、对数似然比（LLR）、加权谱斜率（WSS）和综合指标 CC 和 C。同时，还比较了以下人工耳蜗语音处理技术：传统的 FB、双共振非线性（DRNL）和简单双路径非线性（SPDN）模型。
在整个数据集上，WNN 的平均 SNRseg、MSE 和 LLR 值分别为 2.496±2.794、0.086±0.025 和 2.323±0.281。与其他三种方法（重复测量方差分析（ANOVA）；P<0.05）相比，所提出的方法显著提高了 MSE、SNR、SNRseg、LLR、CC 和 C。在具有 2GB 内存的英特尔双核 2.10GHz CPU 上，对每个辅音或元音，在训练集和测试集上运行该算法（用 Matlab R2013a 编写）的平均时间分别为 9.91±0.87（s）和 0.19±0.01（s）。
所提出的算法准确、精确，因此是传统人工耳蜗的一种有前途的新型插件。虽然调整后的算法相对较快，但需要使用高效的矢量化实现来进行实时人工耳蜗语音信号处理。
