Time-domain aliasing cancellation role in error concealment - Electronics Letters

Introduction: Audio technology has come a long way from magnetic tape recording in the early twentieth century to the current digital music technology. The increase in Internet audio streaming and digital audio broadcasting poses a new challenge, as they are often subject to bandwidth limitation that is incompatible with the high data rate of digital music. However, with the current perceptual audio coding technology, these applications have become a reality and will eventually be further developed to include the wireless medium. MPEG-AAC (advanced audio coder) is the current state of the art in audio compression technology. The CD quality promised at bit rates as low as 64 kbit=s makes AAC a strong candidate for high-quality low-bandwidth audio streaming applications. Each AAC frame is independently decodable. With time-domain aliasing cancellation, the information is carried by two consecutive AAC frames. These features make the scheme favourable in audio streaming applications because it makes AAC more resilient to transmission error. The Internet Streaming Media Alliance (ISMA) has created a transport protocol for streaming MPEG-4 multimedia content over IP networks [1], which is based on the Real Time Transport Protocol (RTP) over UDP (User Datagram Protocol). The IETF Internet Draft regarding RTP payload format for AAC stream can be found in [2] which supports unequal error protection and frame interleaving. Despite this support, handling of erroneous transmission requires a decoder equipped with an error concealment scheme to cater for missing packets. An interleaving strategy was proposed for streaming applications to mitigate the effect of burst error, and especially to avoid consecutive frame error. When adjacent frames are available, the AAC decoder can simply ignore the missing frame, and fill the missing portions with the aliased sample from previous and consecutive frames. This was made possible by the overlap add process in this coder. In this Letter, we illustrate the modified discrete cosine transform (MDCT) with 50% overlap used in MPEG-AAC, and explain how it can be exploited to enhance the concealment scheme in the case of single frame loss. The performance of this approach is reported and compared to both the common muting method for single-frame error concealment and a frame repetition that has similar complexity [3].