Graphical Elicitation Techniques for Subjective Assessment of the Spatial Attributes of Loudspeaker Reproduction – A Pilot Investigation

Non-verbal elicitation techniques may be used in addition to verbal methods in order to obtain meaningful subjective responses about the spatial attributes of reproduced sound. By analysing results from a preliminary graphical investigation, the provision of such responses has been appraised and practical considerations highlighted. Data analysis indicates that non-verbal responses uphold conventional expectations with respect to the effect of loudspeaker and listener location on perceived sound images. With this in mind, it is suggested that the technique be used to assess variables which have not been subject to such intensive study, or be employed in situations where a verbal language may not be appropriate. Further investigations are therefore proposed with respect to the findings of this paper. INTRODUCTION Elicitation experiments have gained favour in subjective audio research as a means of discovering what is actually perceived by a participant when listening to an auditory stimulus. An elicitation phase is therefore a useful starting point for many listening experiments, with the responses obtained from these initial explorations enabling the experimenter to pinpoint more accurately what is perceived, and thus to establish what to evaluate further using subjective assessment. Elicitation techniques have historically used verbal language in order to represent what a listener perceives on hearing a sound. In these experiments, the semantic emphasis is placed on the use of the participant’s own language, rather than on a language provided by an external source, such as the experimenter. This use of individual language can harness the personal constructs of a participant, whilst avoiding the numerous interpretation pitfalls associated with using someone else’s words to describe an event. No matter how useful verbal information is, it is worth remembering that verbal communication is not our only means of communicating, and this is especially true when what is being communicated is not inherently verbal, such as the perception of an auditory event. It may therefore be possible to obtain a greater amount of useful information about what is perceived if additional ‘languages’, including nonverbal response mechanisms, are employed when eliciting percepts from participants in auditory investigations. Not only this, but it may be argued that the greater the similarities between the elicitation method and the percept, the smaller the amount of interpretation required between perception and representation. Less interpretation FORD ET AL. GRAPHICAL ELICITATION TECHNIQUES AES 110 CONVENTION, AMSTERDAM, NETHERLANDS, 2001 MAY 12–15 2 results in more appropriate responses, and, as such, is an important quality for an elicitation method to possess. As non-verbal response techniques appear to provide this close link between response and auditory percept [1], it would seem advantageous to employ such a method in subjective elicitation investigation. With this in mind, the paper describes a preliminary investigation where a simple graphical elicitation technique has been used to obtain individual responses from a small group of musically competent participants. The investigation requires participants to describe, using pencil, tracing paper and their own drawing technique, what they perceive to be the spatial attributes of complex stereo sound sources reproduced over different loudspeaker pairs positioned in three locations within a BS1116 specified listening room. The procedure simultaneously investigates the influence of sub-optimal listener location on the spatial attributes of the perceived image. Although not seeking to prove or disprove a specific hypothesis, the visual responses produced by each listener have been examined using a variety of numerical and graphical techniques to indicate how graphical data can be analysed. Based on this analysis, the paper concludes by suggesting possible applications of the method, and highlights the many practical considerations, exposed during the investigation, that need to be acknowledged prior to the technique’s use in a more rigorous setting. 1. INVESTIGATION CONDITIONS For this preliminary investigation of a graphical elicitation method, it was decided that participants would be asked to respond to complex sound sources reproduced over various loudspeaker locations. Furthermore listeners would be required to move between three listening positions at regular intervals during the presentation of the stimuli. Although using multiple variables added complexity to the investigation, it was felt that a pilot study was the perfect opportunity to experiment on a broad level, enabling the greatest breadth of information to be obtained from the participant, in order that prospective in-depth studies could be identified from within the body of data. In the following sections, the rationale for using the different investigation variables are explained and details of the conditions are outlined. 1.1 Loudspeaker Location Throughout its history, research into loudspeaker sound quality has concluded on numerous occasions (for example [2, 3, 4, 5, 6]) that a loudspeaker’s location can have a direct effect on what is perceived, and likewise what is preferred, by a listener, due to the interaction between loudspeaker and listening room. It is also appreciated that, due to the constraints of the typical domestic environment, it is unlikely that all loudspeakers will be consistent in their placing at the ‘optimum’ locations suggested for the method of audio reproduction. With this obvious conflict of interest between optimum and practically obtainable loudspeaker location, it is undoubtedly important to investigate the influence of position when assessing loudspeaker reproduction for small rooms. And although loudspeaker/room interaction has been covered much in the past, loudspeaker location effects on the perceived spatial attributes of the reproduced image have been less well documented. It is these attributes that are investigated here, using a different method of elicitation to establish initially what kind of data can be elicited which reflects the location of the loudspeaker within the room, and whether these graphical responses correspond with popularly held ideals. Three loudspeaker locations were chosen as variables in this investigation. The first loudspeaker pair, ‘A1’, was located ±30° around the central listening position; seat 2. This loudspeaker location was selected as it is a standard loudspeaker location for the reproduction of two channel stereo material. The ‘A1’ pair was positioned at ear height for an average listener sat in seat 2. The second loudspeaker pair ‘A2’ was positioned, again with tweeter at ear height, ±60° around the central listening location. It was decided that a wider loudspeaker location would be employed as it is more representative of loudspeaker placement in the home environment. Loudspeaker location ‘B3’ was chosen again as a reflection of a more typical consumer loudspeaker positioning, whereby the loudspeakers are removed from the centre of the room to a more practical, if arguably less ‘accurate’ position in the corners of the space. Here the pair was raised to 1.65m above the ground so as to approximate a position on a bookshelf or cabinet. The layout of loudspeakers and listening positions within the room are illustrated in figure 1.