Institute of Acoustics: Paper Detail

Methods for product sound design: a correlation between annoyance and sound quality of noise emitted by household appliances

Samantha Di Loreto 1

Università Politecnica delle Marche Via Brecce Bianche 12, Ancona, Italy

Fabio Serpilli 2

Università Politecnica delle Marche Via Brecce Bianche 12, Ancona, Italy

Valter Lori 3

Università Politecnica delle Marche Via Brecce Bianche 12, Ancona, Italy

ABSTRACT The sound of a product is now a product parameter that needs the same attention as its concept design. However it is often not known how the subjective assessments and the human factor are taken into account in the procedure governing the studies of sound quality. This case study show how the consumer perception of product sound quality is a fundamental variable to define the sound character. Psychoacoustic metrics like Loudness, Sharpness and Roughness, and the combinations of them into more sophisticated models, like annoyance, pleasantness and powerfulness were used for analysis and prediction of kitchen range hoods sound quality. However, some problems arise when the sounds are analyzed by di ff erent consumers. It is assumed that the reason for this is the human not ability to understand the di ff erence between noise and annoyance and to focus consciously or unconsciously on the sound emitted by the appliance. For this reason, the aim of this work is to validate a method for predicting the sound quality produced by five range hoods under reals installation conditions. This was achieved by constructing a series of listening tests suitable for the case study and correlating the sound quality index with perceptual annoyance, to quantify the degree of disturbance obtained in psychoacoustic experiments.

1. INTRODUCTION

It is well known that the Sound Pressure Level (SPL) does not fully represent the acoustic comfort as the parameter “does not take into account the complicated interaction between sounds and human

1 s.diloreto@pm.univpm.it

2 f.serpilli@univpm.it

3 v.lori@univpm.it

a slaty. inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS O ¥, ? GLASGOW

perception of noise” [1]. Compared to the traditional A-weighted Sound Pressure Level, evaluated by sound measurement, the sound quality parameters are better, because they are unanimous with the subjective sensation of the human being [2]. The technical references for this type of measurements are the ISO 532-1 Zwicker-method [3] and ISO 532-2 Moore-Glasberg [4]. In Psychoacoustics Fact and Methods [1], Zwicker describes the characteristics of the human auditory system in acoustic communications because the acoustic communication is one of the fundamental prerequisites for the existence of human society. For the sound quality measurements the psychoacoustic parameters such as Loudness, Sharpness, Roughness and others have been used. The goal of psychoacoustics is implicit, it is to understand how people perceive and experience sound. This can be done at a low level; it may need to understand how accurately a sound source can be detected, or it may need to evaluate the smallest variation in sound level that can be detected. It is possible to have a level up and dare to want to understand the emotional response to sound in the environment in which we live every day, or we may want to evaluate the annoyance due to noise pollution. The result obtained from this approach, still in some ways unknown, is aimed at designing better environments, more attractive products and more e ff ective regulations. Guski [5] proposed that the assessment of acoustic information should be made in two steps: the content analysis of free descriptions of the sound and the uni-dimensional or multidimensional psychophysics. This process requires that di ff erent sounds have to be analyzed. In the first step, the aim is to establish a language suitable for describing the characteristics of the chosen sounds. In the second step, the relation between the verbal descriptions and acoustic metrics may be determined. The aim of this work was developed and validated a method to evaluate the sounds produced by kitchen hoods. This case study is related to a sound quality analysis of some kitchen hood installed in five di ff erent kitchen environments.

2. MATERIALS AND METHODS

Working with sound quality is an iterative process. It starts with the acoustic characterization of prototype of a industrial product, then you make recordings of the sound from your prototypes and at the last you get the first evaluation from a listening test with a jury representing the final users of the product. The fig.[4] shows the iteration process for sound quality approach.

Start

Acoustic analysis

Statistical analysis

Sound’s recordings

Listening Test

Figure 1: Iteration process for sound quality approach

2.1. Acoustic characterization and binaural measurements Identification of sound enables association of the sound to surrounding events. There is no natural relation between form and content of a sign, but it is the interpreter who associates meaning to a form. For this reason the acoustic measurements were carry out in kitchen environments in real condition of functioning of kitchen hood. Five range hoods with varying sound quality are selected, as listed in Tab.1. Binaural recordings of the range hoods was carried out in five kitchen environments with a background noise of between 38 to 42 dBA. The measurement of the sound levels, for the

calculation of the psychoacoustic parameters, were carried out both in an empty kitchen environment and during the lunch time of the diners. This choice was dictated by the need to understand how much environmental noise could a ff ect the subjective evaluation of the hood noise.

Table 1: Selected range hoods.

Type Suction

IEC point suction

Noise max speed

capacity [m3 / h]

Position

Lw [dB (A)]

K1 Top 610 69

K2 Top 550 63

K3 Top 620 65

K4 Bottom 750 62

K5 Bottom 660 66

The sound levels were measured with a head and torso simulator type 4100 with preamplifier type 5953. For the development of the method the psychoacoustic metrics indicated in the reference standard were calculated [3]. The same measurements were made for all kitchen hoods covered by the case study. A representative installation kitchen environment and acoustic binaural measurements are depicted in fig. [2].

Figure 2: Binaural measurements with head and torso simulator in a kitchen environment (K1).

The tab.[2] shows the numerical values of SPL, Loudness, Sharpness, Roughness and Fluctuation Strength for K1 sample at the max speed in empty environment.

Table 2: Results of binaural measurements.

Acoustic and Psychoacoustic

U.M. Right ear Left ear

parameters

Leq dBA 60,2 62,4

Loudness Phone 79,06 79,99

Sharpness Acum 1,91 1,89

Roughness Asper 1,59 1,46

Fluctuation Strength Vacil 0,41 0,43

3. LISTENING TEST

To carry out subjective assessments in the case of systems that generate small impairments, it is necessary to select an appropriate method. In this regard, the advice provided by the ITU Recommendations is an excellent tool. If the jury were not well educated on the test procedure, the obtained results could not be correlated with the psychoacoustic measures and this would lead to the cancellation of the test. For this reason it is important to first define a normalized rating scale which will then be exploited to build the "ad hoc" questions for the case study. Subjective metrics are adjectives that describe the psychological perception that man has of the sound or noise emitted by the product under study. Their choice determines the nature of the sound quality study and therefore the nature of the sound quality index that is in looking for. Test procedures could be realized by single presentations or paired comparisons [6]. The listening tests, in this study, was done in the same environment where the appliance is operating under typical installation conditions. By doing so, the listener will guarantee an extremely subjective opinion also influenced by other sensory aspects. For the subjective evaluation of sound quality a discrete five-degree scale is used where five represents the most score of pleasantness of product and one the less score. The listeners were chosen in the five home where the hoods were installed. To increment the number of listeners the test was proposed, trought a google form, to students and professors from the Università Politecnica delle Marche. The acoustics of room could be a ff ect the headphone playback. Therefore, the playback system was calibrated using a binaural head in controlled acoustic room before to share the record online. The questions were designed specifically for this case study and in particular the aim of questions were to understand how much the listener perceived in terms of pleasantness, noise, annoying and powerfulness of the sound of the product. The fig.[3], shows a sketch of questionnaire.

Figure 3: Questionnaire example

4. CORRELATIONS BETWEEN OBJECTIVE AND SUBJECTIVE METRICS

Models for prediction of perceived annoyance and overall product quality based on acoustic quantities were developed by linear regression [7]. The main goal of this phase of the sound quality process is to identify the most e ff ective correlation between the psychoacoustic parameters and the responses of subjective tests. The purpose of the correlation study is to highlight an interdependence link between the statistical variables. The linear correlation is measured by Bravais-Pearson equation[1]:

r = P n i = 1 ( x i − ¯ x )( y i − ¯ y ) pP n i = 1 ( x i − ¯ x ) 2 p P n i = 1 ( y i − ¯ y ) 2 (1)

Linear regression models are the relation between a dependent, or response, variable y and one or more independent, or predictor, variables x(1),...,x(n). Simple linear regression considers only one independent variable using the relation:

The listener is asked to make a 1-5 rating related of house appliance noise: Pleasantness ‘Noiseless ‘Annoying @00CO Powerfulness @@0@00

Y i = β 0 + β 1 X i + ϵ i (2)

where β 0 is the y-intercept, β 1 is the slope (or regression coe ffi cient), and ϵ i is the error term. It will be start with a set of n observed values of x and y given by:

( x 1 , y 1 ) , ( x 2 , y 2 ) , ..., ( x n , y n ) (3)

Using the simple linear regression relation, these values form a system of linear equations. It is possible to represent these equations in matrix form and then (2) can be expressed more concisely as:

Y = X ∗ B (4)

Taking into account the statistical data, it can be possible to suppose that the sound quality index can be reached through statistical inference methods. The goal is to research the objective metrics to be considered in the representative model of the sound quality index. The search for correlations consists in understanding whether and to what extent a predicted variable, in our case an opinion expressed X, depends (is related) to a measurable characteristic of the sound event judged to be Y (predictive var.) [7, 8]. The parameters of the correlation sought, between X and Y, are the coe ffi cients of the linear regression line eq.[2] and the Pearson index R eq.[1]. The investigated objective metric X is strongly linked to the subjective metric Y and therefore its measurable trend informs us of man’s psychoacoustic perception of noise. Taking into account the statistical data, the correlations were calculated considering the average values between right and left ear. Table [3] summarizes the correlation of Pearson between subjective and objective metrics and shows the results of the polynomial regression for each kitchen hood. The correlations were calculated considering the average values between right and left ear for each samples considered.

Table 3: Values of the coe ffi cients of Pearson for samples K1, K2, K3, K4, K5