Just Noticealbe Difference of Sound Quality Metrics of Refrigerator Noise

A prediction model for the sound quality of refrigerator noise was proposed by investigating subjective and objective attributes of the noise [Jeon et al. (2007) Appl. Acoust.]. In the present study, the just noticeable difference (JND) of each sound quality metric loudness, sharpness, roughness and fluctuation strength which constitute the prediction model was investigated. Loudness of recorded sound samples from five refrigerators was varied according to constant intervals in sound pressure levels. Sharpness was also changed at 14-16 barks in order to provide experimental sounds. Roughness and fluctuation strength were changed by varying amplitude modulation frequencies in time domain. Auditory experiments were conducted to discriminate the JNDs of sound quality metrics by method of limit. The results indicated that JNDs of loudness and sharpness were 0.50 sone and 0.08 acum, respectively. JNDs of roughness and fluctuation strength were also found as 0.03 asper and 0.01 vacil, respectively. These results can be applied to estimate the perceivable difference of sound quality metrics. INTRODUCTION Annoyance of household appliance noise such as refrigerator noise is influenced not only by sound pressure level but also by frequency and time variation of the noise, and also by human auditory characteristics. This is the reason why psychophysical approach should be made as far as sound quality enhancement of appliance noise is concerned [1-2]. In order to investigate sound quality characteristics of refrigerator noise considering real living environments, previous study [3] was done and Zwicker’s loudness, sharpness, roughness and fluctuation strength [4] were found as dominant sound quality metrics describing subjective evaluations. However, just finding and characterizing dominant sound quality metrics of specific sound is not sufficient. In order to improve sound quality of the noise effectively, just noticeable difference (JND) of each metrics should be investigated and also methodologies for the investigation should be defined. Studies on JND are mostly dealt with in the psychoacoustic field, and are mainly concentrated on intensity, frequency and duration discriminations. Pure tone, complex tone, or white noise is applied for the JND studies [5-9] but little is known about the crimination of sound quality metrics using actual noise. Methodologies of JND investigation are classified to method of limit, method of adjustment, and method of constant stimuli classically [10], and simple and transformed up-down methods are also applied to provide unbiased results [11]. Method of limit has advantages of simple procedures of experiments and analyses but requests sufficient pilot-test as its limitations. Whereas, biased results can be reduced by applying transformed up-down method but procedure of the method is complicated comparing that of method of limit. In the present study, JND of sound quality metrics such as loudness, sharpness, roughness and fluctuation strength was investigated applying refrigerator noise and method of limit was used as methodology of discrimination measurements. Experimental sounds were provided by convolution between refrigerator noise recorded in an anechoic chamber and impulse response measured in an actual living room. Value of sound quality metrics of the sounds was varied and auditory tests were conducted to find JND of each of the metrics. JND was determined when more than 75% of subjects participated in the tests showed differential discrimination response