An experimental method for determining factors or frequency ranges that influence product sound quality is described and a method is developed to identify the metrics involved in sound quality or subjective feeling for a product sound. Sound from a vacuum cleaner is used as an example of the use of the proposed techniques and sample sounds were prepared for use in a subjective listening test by increasing or decreasing the sound level of some important frequency bands. The orthogonal array technique was used for designing the experiment or product quality control for determining the relationship between the frequency spectrum and sound quality. Low frequency sounds below the 600 Hz band were closely related to the subjective feeling for the cleaning performance of vacuum cleaners and both low and high (above 3.9 kHz band) frequency components made the sounds less pleasant to hear. In the latter, an attempt was made to extract an "Annoyance Index" for vacuum cleaner sound (AIVC) and the linear regression method and artificial neural networks were employed. The findings indicate that, as the loudness increases, the AIVC value increases abruptly at low loudness values, but the rate of increase slows when a high loudness value is involved and the increasing pattern of the AIVC assumes an S-shape with an increase in sharpness. (C) 2003 Institute of Noise Control Engineering.