Milos Markovic

Three-dimensional point-cloud room model in room acoustics simulations

Telepresence applications require communication with the feeling of being together and sharing the same environment. One important task in these applications is to render the acoustics of the distant room for the telepresence system user. This paper presents a fast method for the room geometry acquisition and its representation with a 3D point-cloud model, as well as utilization of such a model for the room acoustics simulations. A room is scanned with a commercially available input device (Kinect for Xbox360) in two different ways; the first one involves the device placed in the middle of the room and rotated around the vertical axis while for the second one the device is moved within the room. Benefits of both approaches were analyzed. The devices depth sensor provides a set of points in a three-dimensional coordinate system which represents scanned surfaces of the room interior. These data are used to build a 3D point-cloud model of the room. Several models are created to meet requirements of differen...

Beaming teching application: recording techniques for spatial xylophone sound rendering

BEAMING is a telepresence research project aiming at providing a multimodal interaction between two or more participants located at distant locations. One of the BEAMING applications allows a distant teacher to give a xylophone playing lecture to the students. Therefore, rendering of the xylophone played at students location is required at teachers site. This paper presents a comparison of different recording techniques for a spatial xylophone sound rendering. Directivity pattern of the xylophone was measured and spatial properties of the sound field created by a xylophone as a distributed sound source were analyzed. Xylophone recordings were performed using different microphone configurations: one and two-channel recording setups are implemented. Recordings were carried out in standard listening room and in an anechoic chamber. Differences between anechoic and reverberant xylophone sound for binaural synthesis are examined. One-channel recording approach with binaural synthesis for spatial xylophone so...

Evaluation of dynamic binaural reproduction system for live transmitted xylophone recording

For a special teaching application of the telepresence research project BEAMING, a scenario of a remote teacher (the Visitor) teaching a local student to play a xylophone through an embodiment is defined. In order to achieve this, a system is required to record, transmit and render the sound of the xylophone to the teacher in a dynamic scene. In an implementation of such a system, the xylophone is recorded using a mono recording technique. The signal is then processed to spread out the sound of the distributed sound source as multiple point sources in the virtual scene experienced by the Visitor. Finally head tracking allows for a dynamic binaural rendering of the xylophone sound. The goal of this paper is to evaluate the realism of this virtual (auditory) representation of a real xylophone. A listening test is designed to compare characteristics of a real physical xylophone in front of the listener with a rendering using the described system. The evaluation is done with a basis in methods previously used...

Calibration aspects of binaural sound reproduction over insert earphones

Earphones are nowadays widely adopted for the reproduction of audio material in mobile multimedia and communication platforms, e.g. smartphones. Reproduction of high-quality spatial sound on such platforms can dramatically improve their applicability, and since two channels are always available in earphone-based reproduction, binaural reproduction can be applied directly. This paper is concerned with the theoretical and practical aspects relevant to the correct reproduction of binaural signals over insert earphones. To this purpose, a theoretical model originally developed to explain the acoustic transmission to and within the open ear canal is revisited [Mo/ller, Appl. Acoust., 36, 171-218 (1992)]. The model is modified accordingly in order to investigate the aspects of the transmission within the blocked ear canal that are significant to the calibration required to preserve the natural spatial cues that exist during normal hearing conditions, i.e. during an open-ear-canal situation. To evaluate the vali...

Improvement in design of noise shaping filters

Addition of dither before the quantization (or requantization) linearizes the quantization characteristic for signals of low levels. However, the dither addition increases overall noise, which is uniformly distributed within the whole frequency range. By transposing a part of that energy into frequency regions where a human ear is less sensitive, the present noise can become less audible. Therefore, noise shaping is incorporated in the quantization (requantization). A noise shaping filter can be designed by applying suitable weighting functions that represent audibility to low level noise. The filter’s power spectral density should approximate the inverse of the audibility curve. In that way, perceived signal to noise ratio becomes higher. In this paper, some aspects of the noise shaping filter design are analyzed, and an improvement of the design procedure is proposed. The improvement is related to better matching of the filter’s power spectral density to the corresponding target curve. The results obtained by the proposed procedure are verified by listening tests.