Part III: Measuring hearing-critical job tasks 119 Statistical analyses The acoustical measurements were analysed in inMatlab with the Statistical Toolbox (The Mathworks, 2005) and with Statistica (StatSoft, 2009). Descriptive statistics were calculated for the level and spectrum of the warning signals and the noise field, as well as for the laboratory measurements. For all train types, octave band spectra in dB SPL as well as the A-weighted equivalent sound pressure level in dBA (LAeq) were presented when driving 80 km/h, 100 km/h and when driving at maximum speed. Differences in ambient noise levels between the train types were investigated using a repeated measures analysis with train type, driving speed, and the interaction between train type and driving speed as independent variables. This analysis was performed with the Statistical Package for Social Sciences (SPSS) version 25.0 (Armonk New York USA). The detectability of the DSD and ATP signal was estimated with the Detectsound software (Zheng et al., 2003) that has been developed to evaluate the efficacy of auditory warning signals in noisy workplaces (Laroche et al., 1991). Using this software, the detectability of acoustic warning signals in real-life conditions can be predicted. A validation study has revealed that the mean error in estimating detection thresholds in continuous noise fields is typically within 1 dB with a standard deviation of less than 2.5 dB (Zheng et al., 2007). To run Detectsound, the acoustic characteristics of the ambient noise at the workplaces and the warning signals should be obtained in 1/3 octave band levels from 125 to 12500 Hz (Laroche et al., 1991). Warning-signal detectability is predicted for each workplace by comparing the spectral content of the warning signal with the predicted optimal range, also known as ‘the design window.’ Therefore, the masked detection threshold is calculated according to the acoustical characteristics of the noise and the hearing status of the receivers (Proulx et al., 1996; Zheng et al., 2003). In ISO 7731, a signal level is proposed of 10–15 dB above themasked threshold, and a warning signal is advised to have signal components in the range of 500–2500 Hz (ISO, 2003). In line with these recommendations, the lower and upper limit of the design window are respectively 12 and 25 dB above the masked detection threshold for the detection of the warning sound in the given noise field in frequencies ranging from 125 to 3150 Hz (Zheng et al., 2003). In ISO 7731, it is proposed that at least one spectral component should reach the design window (ISO, 2003), but several authors have suggested that more than one component is required to account for the common fluctuations in
RkJQdWJsaXNoZXIy MTk4NDMw