Subjective Evaluation of Loudness Models Using Synthesized and Environmental Sounds

Loudness of different sounds was measured in order to evaluate Zwicker’s and Moore’s loudness models (published in Acta Acustica [3] and in the J. Audio. Eng. Soc. [4]). The test signals were synthesized bands of noise and environmental sounds. For synthesized noises, Zwicker’s model fits highly well with subjective results, it is a little bit closer to the data than Moore’s. For environmental sounds, calculated values from Moore’s model fit very well with subjective results for low-level signals. In this case, Zwicker’s model is close to the subjective data for loud signals. 1 INTRODUCTION In 1965, Zwicker and Scharf proposed a model for calculating the loudness of complex sounds [1]. A BASIC-Program of that model (usually called Zwicker’s model) has been published [2]; we used it to develop our own program. This model has recently been revised by Moore and Glasberg [3, 4]. Zwicker’s model have been evaluated for bands of noise [1] and for environmental sounds [5]. Moore et al. compared their model to Zwicker’s for bands of noise [3, 4]. The aim of this paper is to compare the loudness calculated using the two models with loudness estimated by listeners both for synthesized noises and for environmental sounds. 2 EXPERIMENT 1: LOUDNESS OF SYNTHESIZED NOISES 2.1 Method In a first experiment, the loudness of synthesized bands of noise has been evaluated. Three different center frequencies were chosen. For each center frequency, we selected two different bandwidths and for each bandwidth two sound pressure levels. Altogether, twelve bands of noise have been used (Table 1). Eight listeners with normal hearing participated in the experiment. Their age ranged from 20 to 29 years, and their thresholds were 10 dB HL or lower for frequencies from 0.5 to 8 kHz (BK 1800 audiometer). The task of the listener was to make loudness matches between bands of noise. For each run, the reference noise had a fixed level, and the comparison sound (called the test sound) was varied in level to achieve equal loudness. The sound durations were 1 s with a 500-ms interstimulus interval. sound number Central frequency (Hz) bandwidth (Hz) Level (dB SPL) sound number Central frequency (Hz) bandwidth (Hz) Level (dB SPL) 1 400 50 40 7 1420 120 50 2 400 420 40 8 142