Spatial Design

The BPK Experience

There are three distinct parties involved in a performance with the BPK: the audience, the users, and the programmers. Each of them has a different vantage point. The audience are predominantly people of public spaces. They observe two persons being physically and operatively connected to one unit. Movement and speech appears scripted and exhibits a recognizable coherence. Stories are told by two people in unison and a mysticism is generated regarding the internal workings of the box. A private sphere of knowledge is overtly prominent within the public space.

The users wear the device like a helmet. They may roam spaces freely as long as they accommodate for their physical link. Two separate audio feeds--one for each user--tell stories as to why it is important to adhere to the instruction that are to follow. This sets the tone for the users' performance. Once properly introduced the device starts cuing movements and utterances ("move two steps forward, proclaim that your partner is rather slow today"). By sensing physical acceleration and analyzing spoken words the system gauges the level of participation and adjusts the cuing rate accordingly. Should the users discontinue their involvement the device falls back into persuasion mode explaining the importance of this setup. As an additional incentive to engage with the device the users will hear an increasingly more articulated percussion line as they intensify their participation.

The programmers of the box deal with a platform for experiments in the field of human-computer interaction. Unlike the usual one-to-one relation the BPK provides a triangular setup. Part of this setup are exclusive communication channels between the BPK and either one of the participants. This gives the programmers control over four different spheres of knowledge. They can specifically direct story lines at either participant separately, both participants, or the audience. By telling different stories to separate parties they can cause variance in the conception as to what the performance is all about, provoke interference, and interpersonal dynamics.

Implementation

The primary elements are software, electronics, and enclosure. The software runs on an embedded Linux system which is mounted inside the rectangular box that defines the physical appearance of the BPK. The box is made from opaque white polystyrene. This material has a matt homogeneous finish, is durable, and is almost as easy to work with as cardboard. To cut out pieces one simple has to scribe the sheets with a knife and carry out a snap motion. The particular pieces are assembled by epoxy-gluing polycarbonate angles into the corners of the box. The box is mostly static. The only kinetic elements are two lazy-susan-like discs on the internal side of the top sheet where the users' heads touch the structure. This allows hassle-free wearing of the device. Similar to a cap, one simple places it on top of one's head. The rotating discs between the head and the box assure free movement around one's vertical axis.

The electronics part is based on two embedded PC boards from the Via Epia series. Each of them is approximately the size of two CD covers, is binary compatible to the x86 architecture, and provides all the standard PC inputs/outputs. Connected via fast-ethernet they provide six input and four output audio channels. Three output channels are used to play audio feeds for each user and the "unifying soundscape." Two input channels are used for voice-analyzing the spoken words of the users. For three degree of freedom acceleration sensing the BPK utilizes the SerAccel module from Sparkfun. This module plugs into the serial port and can be read out with any application that has access to this port.

The control software is a combination of FOSS (free open source software) programs and custom Python code. The BPK uses Debian Linux, the Sphinx2 speech recognition engine, RTcmix, and the Python interpreter. The general flow is based on a simple Markov chain in which each jump from a series of intuitively composed pro-scripted events to the next is weighted by sensor input. For example, low-frequency sound synthesis is favored by backward motion.