Proceedings of the Stockholm Music Acoustics Conference, August 6-9, 2003 (SMAC 03), Stockholm, Sweden HARDNESS RECOGNITION IN SYNTHETIC SOUNDS Bruno L. Giordano, Karin Petrini Department of General Psychology, University of Padova, Italy bruno.giordano@unipd.it ABSTRACT Sound source recognition investigates recovery of different fea- tures of the objects, whose interaction lead to the generation of the acoustical signal. Among them material type have received par- ticular attention, while recovering of material properties, such as hardness, have been scarcely considered. Hardness plays a signif- icant role in the musical field too, especially for percussion instru- ments, where resonating objects of variable hardness are struck with mallets of variable hardness. Comparison of previous results on hardness recognition point toward the perceptual independence of the resonator and exciter properties. This issue was addressed in four experiments conducted on stimuli synthesized with a physical model, which allowed independent manipulation of the exciter and resonator properties. Free identification and forced choice tasks have been used to investigate the ability of listeners to discriminate variations in the exciter from variations in the resonator. Scaling tasks have been used to investigate the relationship between the synthesis parameters and the hardness estimates of the exciter and of the resonator. Free identification and forced choice data reveal a bias toward the interpretation of the acoustical signals in terms of features of the resonating object. Hardness scaling results reveal the perceptual dependence of exciter and resonator properties, al- though strong individual differences are found. 1. INTRODUCTION A relatively recent field in auditory perception investigates recog- nition of the features of the sounds source [1]. The most stud- ied class of signals is that of impact sounds [2], [3], generated by the interaction between a highly damped object, the exciter or hammer, and a vibrating object, called resonator or sounding ob- ject. Particular interest have received the study of categorization or discrimination of the material type of the resonator [4], [5], [6], [7]. Wildes and Richards [8] hypothesized that material could be uniquely specified, at the acoustical level, by means of the tanφ coefficient, which measures the degree of anelasticity of the mate- rial of the resonator, defined as tanφ =1/πfte (1) where f is the frequency of vibration and te is the time it takes for the amplitude to decay to 1/e of its starting value. Different perceptual studies found material categorization to be strongly in- fluenced by this acoustical parameter (rubber and wood chosen for higher tanφ values than glass and steel)[5], [4]. Other acoustical parameters, modulated by physical features extraneous to the ma- terial type of the resonator, were found to influence recognition of this sound source feature: the decay time of signal amplitude, as modulated by external damping of the resonator, and signal fre- quency, as modulated by the size of the resonator [6]. In a fur- ther experiment Giordano [6] demonstrated recognition of material type of the resonator to be independent of the material of the ex- citer, although the range of variation in the elastic properties of the used exciters was too low to induce a relevant timbral variation in the generated signals. Scaling of a mechanical material property, namely mallet hardness, was investigated by Freed [9]. Different cooking pans of variables size were struck with mallets of variable hardness. Participants were found to properly scale the hardness of the mallets, independently of the size of the resonator. Hardness estimates were found to be influenced by signal amplitude, spectral centroid, and amplitude decay velocity. On the basis of these re- sults, recognition of the material properties of the resonator can be hypothesized to be independent of the features of the exciter, and viceversa. This was first assessed by investigating the ability of listeners to distinguish variations in the resonator from variations in the exciter features. Both free identification and forced choice tasks were used. Second, scaling of the hardness of both the exciter and the resonator was conducted, in synthetic stimuli generated by manipulating the properties of the resonator as well as those of the exciter. These topics are of interest to the musical field too, par- ticularly for the field of percussion instruments, where resonators made of a large variety of materials (e.g., metallic and wooden instruments, membranophones, lithophones, glass and ceramic in- struments) are struck with objects of variable hardness (e.g., felt and wooden mallets, metallic sticks, hands) [10]. Then research on hardness scaling could partially highlight the problem of the perceptual quality of percussion instruments, scarcely considered in timbre perception research. 2. SYNTHESIS PARAMETERS Stimuli were generated with the impact model by Avanzini et al. [11]. Three parameters of the model were investigated. The first two were related to the resonator: the frequency of the lowest mode f , and τ =1/πtanφ. The first parameter can be thought as modelling the geometrical properties of the resonator (e.g., length). The τ parameter models the material of the resonator. High τ val- ues correspond to hard or stiff materials (steel, glass). Low τ val- ues correspond to soft or elastic materials, such as rubber. The third synthesis parameter was related to the interaction between the exciter and the resonator: the force stiffness coefficient k, di- rectly related to the elastic properties of the exciter. Low k values may represent rubber or felt mallets. High k values may represent stiff mallets (metal, hard woods). 3. FREE IDENTIFICATION A free identification experiment was conducted in order to assess how variations in the resonator or in the exciter parameters were described or interpreted by listeners. A set of 16 stimuli was syn- 611