Journal of Medical and Biological Engineering, 33(6): 576-582 576 Objective Measurement of Speech Quality for Hearing Aids Ying-Hui Lai 1 Shih-Tsang Tang 2 Kuen-Shian Tsai 1 Hsiu-Wen Chang 3 Shuenn-Tsong Young 4,* Woei-Chyn Chu 1 1 Hearing and Speech Engineering Laboratory, Institute of Biomedical Engineering, National Yang-Ming University, Taipei 112, Taiwan, ROC 2 Department of Biomedical Engineering, Ming Chuan University, Taoyuan 333, Taiwan, ROC 3 Department of Speech and Hearing Disorders and Sciences, National Taipei University of Nursing and Health Sciences, Taipei 112, Taiwan, ROC 4 Holistic Education Center, Mackay Medical College, New Taipei City 252, Taiwan, ROC Received 29 Nov 2012; Accepted 7 Mar 2013; doi: 10.5405/jmbe.1383 Abstract Hearing aids are commonly used to overcome impairment associated with hearing loss. Recent surveys have shown that hearing aid users are often dissatisfied with the speech quality. However, only subjective methods (e.g., questionnaires) are used to assess speech quality, and the objective methods used to verify the performance of hearing aids have major limitations. This study proposes an objective method called variations of dynamic range (VDR) score for assessing hearing aid speech quality. The method uses quantitative differences between spectra for aided and unaided speech based on Speechmap. Two experiments were performed in this study: (1) one to verify the correlation between scores for the perceptual evaluation of speech quality (PESQ) and VDR in assessing speech quality produced under three compression ratios (CR) of wide-dynamic-range amplification in one channel, and (2) one to verify that the VDR score is consistent with results from clinical trials in the literature. The experiments involved hearing aids fitted using desired sensation level [input/output] (DSL [i/o]) and adaptive dynamic range optimization (ADRO), and stimulated by speech sounds at moderate and loud levels. The VDR scores were measured using the Audioscan Verifit hearing aid fitting system to assess the speech quality of hearing aids. The results indicate that PESQ scores were 3.56, 2.25, and 1.98, and VDR scores were 2.5, 29, and 36 dB for CRs of 1.0, 2.0, and 4.0, respectively. The VDR and PESQ methods provided similar assessments of speech quality produced by hearing aids fitted for flat mild hearing loss (high and low PESQ and VDR scores, respectively, indicate good speech quality). In addition, the average VDR scores were significantly lower (p < 0.05) for hearing aids fitted using ADRO than those for hearing aids fitted using DSL [i/o]. The VDR scores are consistent with PESQ scores for hearing aids with different prescriptions, and are consistent with clinical reports that found the speech quality for hearing aids fitted using ADRO to be higher than that obtained using DSL [i/o]. These results indicate the potential of the VDR score for accurately assessing the speech quality of hearing aids. Keywords: Hearing aids, Speech quality, Speechmap, Dynamic range compression, Adaptive dynamic range optimization (ADRO) 1. Introduction Hearing loss is a relatively common condition (especially among the elderly) that can decrease the quality of life in various ways, including making it difficult to enjoy music and communicate with other people. Sensorineural hearing loss (SNHL) is the most common type of hearing loss [1], which is usually manifested as a reduced sensitivity to soft sounds, with loud sounds producing discomfort [2]. The adverse effects of SNHL can be improved by the use of a hearing aid. However, the characteristics of SNHL are diverse and differ markedly * Corresponding author: Shuenn-Tsong Young Tel: +886-2-26360303 E-mail: young0210@gmail.com between individuals, which has led to the development of numerous hearing aid fitting strategies (so-called prescriptions) to ensure that the processed sounds are suitable for an individual hearing aid wearer with SNHL [3,4]. Wide dynamic range compression (WDRC) is a common amplification scheme in modern hearing aids. It is a type of nonlinear amplification that applies different gains to different levels and frequencies. The WDRC fitting strategy is designed to make low-level signals audible while maintaining comfort for higher-level inputs [3,5,6]. Several commonly used independent prescriptions are based on WDRC, including NAL-NL1 (National Acoustic Laboratories-Nonlinear Version 1) [3] and DSL input/output (DSL [i/o]) [7]. However, WDRC does not provide high-quality speech, as confirmed by Kochkin [8], who reported that 88.4% of hearing aid users considered it