Perception of tonalness of tyre/road noise and objective correlates

Transcription

1 The 33 rd International Congress and Exposition on Noise Control Engineering Perception of tonalness of tyre/road noise and objective correlates S. Buss, R. Weber Oldenburg University, Faculty of Natural Sciences, Institute of Physics, Acoustics Group, Germany Abstract [256] The geometry of the tread pattern has an important influence on tyre/road noise. This pattern noise is a tonal component of the tyre/road noise with speed dependent frequency. If tyre/road noise is perceived inside the car, this leads to a decreased quality rating in a subjective evaluation of tyre/road noise. In order to find an objective parameter describing the subjectively perceived pattern noise, tyre/road noises are evaluated in subjective tests. The evaluated tyre/road noises are tyre/road noise recordings and modifications of these signals with systematically varied spectra. For the subjective assessment of pattern noise correlating objective signal parameters will be presented. Also objective parameters calculated after a preprocessing with an aurally adequate signal analysis are compared to the subjective evaluation of the tyre/road noises. 1 INTRODUCTION While a car is rolling pattern noise is generated at the contact area between the tyre and the road surface. The pattern noise has speed dependent frequency components around the 6th order of the tyre revolutions according to the distribution and the number of the tread pattern elements. This range is called the first pitch harmonics. An increasing level of the first pitch harmonics leads to an increased pattern noise strength [1]. In order to find an objective parameter describing the subjective perception of pattern noise strength, three experiments are carried out. The subjective rating obtained in the experiments will be correlated to two objective level parameters. 2 SUBJECTIVE EVALUATION OF PATTERN NOISE Three experiments are carried out concerning the subjective evaluation of pattern noise: - Experiment 1 (Level variation of the first pitch harmonics): 84 signals with a systematically varied level of the first pitch harmonics are rank ordered according to their pattern noise strength (see also [1]). The first pitch harmonics of four tyre/road noise recordings is amplified by between -1 and 1 db. With a 2AFC-procedure for different tyre/road noises (original recording or a modification) the amplification of the first pitch harmonics of 1/8

2 another recording is measured leading to the same pattern noise strength. The experiment is carried out by four subjects: one female and three male. - Experiment 2 (Variation of the spectral shape of the first pitch harmonics): In a paired comparison test Thurstonian scale values according to the pattern noise strength of 6 signals with a systematically varied spectral shape of the first pitch harmonics are obtained. The signals are based on one tyre/road noise recording. Besides the original spectral shape the different spectral shapes are obtained by an amplification of one or two third around the mean of the first pitch harmonics by 2 or 4 db and a following attenuation of the level of the first pitch harmonics to the original level. Since the differences between these signals is small a paired comparison test is carried in order to have a great accuracy of the subjective rating. By means of Thurstone's law of comparative judgment the subjective ratings are interval scaled. Eight subjects take part in the experiment: two female and six male. - Experiment 3 (Variation of the level and the spectral shape of the first pitch harmonics): In the first two experiments the influence of the level of the first pitch harmonics and the spectral shape on the subjective evaluation of the pattern noise strength are assessed isolated from each other. In a third experiment the combined influence of both parameters on the pattern noise strength is evaluated. A second paired comparison test leads to Thurstonian scale values according to the pattern noise strength of 12 signals with a systematically varied level and spectral shape of the first pitch harmonics. All signals are based on one tyre/road noise recording. Here the level of the first pitch harmonics is the original level and an amplification of the first pitch harmonics by.5 and 1 db. The different spectral shapes are the original shape and an amplification of one third around the mean of the first pitch harmonics by 2 or 4 db and an amplification of two third around the mean of the first pitch harmonics by 4 db. The first pitch harmonics than is attenuated in order to leave the level of the first pitch harmonics constant for all different spectral shapes of the first pitch harmonics. The experiment is carried out by ten subjects: three female and seven male. The assessed signals are tyre/road noise recordings with an artificial head standing on the front passenger seat while the car is coasting and modifications of these signals. The signals are about 6s long and represent a coasting from 11 to 1 km/h. The experiment is carried out in a soundproof chamber. The signals are played back via headphones. All signals have the same db(a)-level of 69 db(a). The subjects are untrained in evaluating tyre/road noise. 3 OBJECTIVE SIGNAL PARAMETERS 3.1 Level of the first pitch harmonics From the averaged order spectrum of the left and right channel the level in the range of the first pitch harmonics (approx. 5th to 85th order) is calculated. The subjective ranking of the signals in the first experiment is significantly correlated with the level of the first pitch harmonics. This can be seen in figure 1 for one subject (see also [1]). The correlation coefficients of all subjects are between.85 and.92 and are therefore significant. 2/8

3 8 ranking according to the pattern noise strength level of the first pitch harmonics [db(a)] Figure 1: Ranking of the signals according to the pattern noise strength in the first experiment (variation of the level of the first pitch harmonics) in dependence of the level of the first pitch harmonics. This plot shows a typical ranking by one subject. There is a significant correlation (r=.92*)(* = significant with p.5) Thurstonian scale value accoring to the pattern noise strength level of the first pitch harmonics [db(a)] Figure 2: Interval scaled Thurstonian scale values of the signals representing the pattern noise strength in the second experiment (variation of the spectral shape of the first pitch harmonics) in dependence of the level of the first pitch harmonics. There is no significant correlation (r=-.8). 3/8

4 The paired comparison data of the experiment concerned with the variation of the spectral shape of the first pitch harmonics is interval scaled by Thurstone's law of comparative judgment. These interval scaled data is plotted against the level of the first pitch harmonics. As shown in figure 2 the Thurstonian scale values of the signals in the second experiment do not correlate with the level of the first pitch harmonics. In this experiment the spectral shape of the first pitch harmonics of the signals is varied while the level of the first pitch harmonics is held nearly constant. It can be seen that the level of the first pitch harmonics can not be the only parameter being related to the pattern noise strength. Also the spectral shape has an influence on the amount of perceived pattern noise. 3 Thurstonian scale value according to the pattern noise strength level of the first pitch harmonics [db(a)] Figure 3: Interval scaled Thurstonian scale values of the signals representing the pattern noise strength in the third experiment (variation of the level and the spectral shape of the first pitch harmonics) in dependence of the level of the first pitch harmonics. There is no significant correlation (r=-.23). In the third experiment there is also no correlation between the Thurstonian scale values and the level of the first pitch harmonics (figure 3). Since the level and the spectral shape of the first pitch harmonics seem to have an influence on the perceived pattern noise strength, it can not be described by the level of the first pitch harmonics alone. 3.2 Level of the first pitch harmonics of the frequency tracks The frequency tracks represent the tonal components of the tyre/road noise. By means of the software VIPER (Cortex Instruments) the frequency tracks of the signals are calculated. The procedure is based on the aurally adequate signal analysis by calculating contours in [2] and [bib4]. The time course of tonal component characterised by local maxima in level at each point in time are extracted. Also the masking properties of the human ear are regarded. The noisy parts of the signal are not further analysed. From the frequency tracks an order spectrum is calculated. Analog to the calculation of the level of the first pitch harmonics the level of the first pitch harmonics of the frequency tracks is calculated as the level of the first pitch harmonics from the order spectrum of the contours. By this value the level of the tonal components of the first pitch harmonics is represented. 4/8

5 8 ranking according to the pattern noise strength level of the first pitch harmonics of the frequency tracks [db] Figure 4: Ranking of the signals according to the pattern noise strength in the first experiment (variation of the level of the first pitch harmonics) in dependence of the level of the first pitch harmonics of the frequency tracks. This plot shows a typical ranking by one subject. There is a significant correlation (r=.96*)(* means significant with p.5). The ranking of the signals in the first experiment is significantly correlated with the level of the first pitch harmonics of the frequency tracks (figure 4). For all subjects the correlation coefficient is significant and lies in the range.94 to.97. This correlation is stronger than in the case of the level of the first pitch harmonics. This finding indicates the tonal components of the first pitch harmonics being the important factor of the pattern noise. Since the pattern noise is a tonal component of the tyre/road noise, this result could be expected. Though there is a general increase of the Thurstonian scale values of the second experiment with the level of the first pitch harmonics of the frequency tracks, the correlation is not significant (figure 5). But in contrast to the level of the first pitch harmonics there is a clear tendency of an increase of pattern noise strength with increasing level of the first pitch harmonics of the frequency tracks. So also in this experiment an increasing pattern noise strength with increasing level of the first pitch harmonics of the frequency tracks can be assumed, but it does not get significant. 5/8

6 3 2.5 Thurstonian scale value according to the pattern noise strength level of the first pitch harmonics of the frequency tracks [db] Figure 5: Interval scaled Thurstonian scale values of the signals representing the pattern noise strength in the second experiment (variation of the spectral shape of the first pitch harmonics) in dependence of the level of the first pitch harmonics of the frequency tracks. There is no significant correlation (r=.46) 3 Thurstonian scale value according to the pattern noise stregth level of the first pitch harmonics of the frequency tracks [db] Figure 6: Thurstonian scale values of the signals representing pattern noise strength in the third experiment (variation of level and spectral shape of the first pitch harmonics) in dependence of the level of the first pitch harmonics of the frequency tracks. There is a significant correlation (r=.91*)(* = significant with p.5). 6/8

7 If the level and the spectral shape of the first pitch harmonics are varied (third experiment), the Thurstonian scale values are significantly correlated with the level of the first pitch harmonics of the frequency tracks (figure 6). So it can be concluded that the level of the first pitch harmonics of the frequency tracks describes the influence of the level as well as the spectral shape of the first pitch harmonics on the perceived pattern noise strength. 3.3 Comparison between the order spectra of the original signal and the frequency tracks 55 5 original frequency tracks 45 4 level [db(a)]/level [db] order Figure 7: Order spectra of one signal: order spectrum of the original signal (solid line) and order spectrum of the frequency tracks (dashed line). In order to demonstrate the influence of the calculation of frequency tracks on the signal, as an example the order spectra of one signal used in the experiments are plotted in figure 7. The solid line represents the order spectrum of the original signal. After calculating the frequency tracks the order spectrum looks like the dashed line. The level of the order spectrum of the frequency tracks is lower than the level of the order spectrum of the original signal. This level difference is a result of the reduction of the signal to the tonal components by deleting the noisy parts of the signal. Another difference between both order spectra is the shape of the first pitch harmonics. The order spectrum of the original signal shows a broad level increase in the range of the first pitch harmonics ranging from about the 45th to about the 9th order. In the order spectrum of the frequency tracks the first pitch harmonics has a clear level increase in the range from about the 5th to about the 7th order. So by calculating frequency tracks the first pitch harmonics is better distinguished from other parts of the signal. 7/8

8 4 CONCLUSION Three experiments are carried out in order to evaluate the influence of the level of the first pitch harmonics and the spectral shape of the first pitch harmonics and the combined influence on the perceived pattern noise strength. The level as well as the spectral shape have an influence on the perceived pattern noise strength. Two objective parameters are related to the subjective evaluation of the pattern noise strength: a) the level of the first pitch harmonics and b) the level of the first pitch harmonics of the frequency tracks representing the level of the tonal components of the first pitch harmonics. If only the level of the first pitch harmonics is varied (Experiment 1), the pattern noise strength increases with increasing level of the first pitch harmonics. If as well or only the spectral shape is varied (Experiment 2 and 3), the level of the first pitch harmonics is not related to the pattern noise strength. Whereas an increasing level of the first pitch harmonics of the frequency tracks and therefore an increasing level of the tonal components of the first pitch harmonics lead to an increasing pattern noise strength. This relation can be found for all three experiments. It gets significant in the case of a level variation of the first pitch harmonics (Experiment 1) and also an additional variation of the spectral shape of the first pitch harmonics (Experiment 3). So with increasing level of the first pitch harmonics of the frequency tracks the pattern noise strength increases. The level of the first pitch harmonics correlates with the rated pattern noise strength only, if the spectral shape is held constant. So the level of the first pitch harmonics of the frequency tracks shows a stronger relation with the pattern noise strength than the level of the first pitch harmonics. Both parameters differ only in the calculation of frequency tracks before calculating the level of the first pitch harmonics. It is shown that an aurally adequate signal analysis extracting tonal components increases the correlation between objective signal parameters and the subjective evaluation of pattern noise. This parameter takes into account both the level and the spectral shape of the first pitch harmonics of the first pitch harmonics. REFERENCES [1] S. Buss, R. Weber and W. Liederer, ``Objektivierung des subjektiv wahrgenommenen Profilgeräuschs in Reifen- Fahrbahn-Geräuschen'', in Fortschritte der Akustik - DAGA 22, Bochum, 22. [2] M. Mummert, Sprachkodierung durch Konturierung eines gehörangepaßten Spektrogramms und ihre Anwendung zur Datenreduktion, VDI-Fortschrittberichte, Reihe 1, VDI-Verlag, Düsseldorf [3] S. Buss and R. Weber, ``Subjective and objective characterization of tonal components in tyre/road noise'', in Fortschritte der Akustik - DAGA 24, Strasbourg, 24. [4] P. Daniel, ``Auditory Spectrograms and Auditory Contours in Musical Acoustics'', in Fortschritte der Akustik DAGA 24, Strasbourg, 24. 8/8