181 An Error – Concealment Technique for Wireless Digital Audio Delivery Nicolas Tatlas 1 , Andreas Floros 2 Thomas Zarouchas 1 and John Mourjopoulos 1 1 AudioGroup, WCL, Univeristy of Patras 265 00 Rio, Patras, Greece {ntatlas, thozar, mourjop}@wcl.ee.upatras.gr 2 Dept. of Informatics, Ionian University, Plateia Tsirigoti 7, 49 100 Corfu, Greece floros@ionio.gr Abstract – Real-time digital audio streaming over unreliable wireless IP networks, such as WLANs, encounters playback quality degradation, due to late arrivals or permanent loss of the transmitted packets. In this paper we present a perceptually efficient error concealment technique, which, combined with playback synchronization correction schemes and a high-rate wireless transmission protocol with Quality- of-Service support, can significantly improve the overall audio playback quality of typical multichanel home-theater wireless applications. I. INTRODUCTION During the last decades, the audio technology has met a significant evolution from analogue to high-quality digital applications. The common element in all typical audio systems is that the equipment interconnection is performed through cables that realize physical wired links carrying the audio information. With the recent developments on digital signal representation, storage and processing leading the shift to an all-digital audio reproduction chain, the audio information is mainly distributed in digital formats through synchronous, wired digital interfaces (e.g. SPDIF, FireWire, HDMI etc) and lately asynchronously over packet-based networks and protocols such as the Internet Protocol (IP). IP-based audio delivery provides the ability to establish numerous audio playback setups that are impossible to realize using conventional, sequential audio systems interconnections, such as music distribution from any digital audio source within a typical multiroom home environment to any networked speaker/playback device. In practice, each IP-enabled audio device can be connected to every other, using the common IP network interface also used for computer communications and Internet access [6]. Nowadays, the IP network interface is frequently wireless. This further extends the supported applications by providing flexible, portable, wire-free and low cost means of packet-based communications. With the recent advances of the Wireless Local Area Networks (WLAN) technology on the offered bandwidth capacity (the recently ratified 802.11g [1] specification offers data rates in the range of 54Mbps, while the forthcoming 802.11n protocol is expected to increase this rate up to 480Mbps), a new set of cable-free, high-quality multimedia applications can be efficiently supported, such as wireless multichannel audio playback for typical home-theater setups. In order to achieve high-quality digital audio wireless playback, a number of issues related to the wireless interconnection nature must be resolved, such as the minimization of the audible distortions introduced during the wireless transmissions and the absolute clock and packet playout synchronization among the multiple audio wireless devices. Previously published works [2], [3], [6] have introduced novel efficient methods for compensating the effect of hardware clock and packet-based playback synchronization. However, depending on the wireless link quality and in order to maintain the relative receivers’ playback synchronization, a significant number of silence gaps are introduced, due to the permanent loss or excessive delay of the transmitted data packets. The motivation for this work was the development of a technique for vastly improving the audible effect of temporal playback muting during the wireless delivery of digital audio data. One of the major initial design requirements of the proposed methodology (termed as Fading Pattern Repetition - FPR) was the low implementation and computational complexity, as it should be applied in real-time prior the final audio playback. As it will be explained later in the paper, this requirement was achieved by a perceptually efficient error concealment algorithm that repeats appropriately processed, successfully received audio data packets. The rest of the paper is organized as follows: Section II presents a brief overview of the technology issues related with the digital audio delivery over WLANs, while Section III analyzes and outlines the implementation of the proposed error concealment technique. In Section IV, the testing procedures and criteria for evaluating the perceptual performance of the audio playback are described, followed by the results obtained during the tests and the conclusions summarized in Section V. II. OVERVIEW OF AUDIO OVER WLANS In general, two different types of wireless home audio applications exist, defined in [3] as: a) point-to-point home audio delivery, were a device acts as the “media server” (possibly integrating all available audio sources, including Internet-stored media) that wirelessly streams the same (or different) audio content to a number of wireless audio players and b) point-to-multipoint delivery, consisting of a multichannel digital audio source transmitting audio to multiples receivers (e.g. DVD audio content to 5.1 wirelessly connected loudspeakers) that perform simultaneous and synchronized playback. In both cases, wireless audio transmission can be theoretically performed using a variety of standardized or proprietary protocols, starting from low-quality analog systems operating in the range of 800-900MHz and up to high-rate digital streaming protocols operating in the