An error-protected speech recognition system for wireless communications

Abstract

uture wireless multimedia terminals will have a variety of applications that require speech recognition capabilities.In this paper, we consider a robust distributed speech recognition system where representative parameters of the speech signal are extracted at the wireless terminal and transmitted to a centralized automatic speech recognition (ASR) server. We propose several unequal error protection schemes for the ASR bit stream and demonstrate the satisfactory performance of these schemes for typical wireless cellular channels. In addition, a "soft-feature" error concealment strategy is introduced at the ASR server that uses "soft-outputs" from the channel decoder to compute the marginal distribution of only the reliable features during likelihood computation at the speech recognizer. This soft-feature error concealment techniques reduces the ASR error rate by up to four times for certain channels. Also considered is a channel decoding technique with source information that improves ASR performance. uture wireless multimedia terminals will have a variety of applications that require speech recognition capabilities. In this paper, we consider a robust distributed speech recognition system where representative parameters of the speech signal are extracted at the wireless terminal and transmitted to a centralized automatic speech recognition (ASR) server. We propose several unequal error protection schemes for the ASR bit stream and demonstrate the satisfactory performance of these schemes for typical wireless cellular channels. In addition, a "soft-feature" error concealment strategy is introduced at the ASR server that uses "soft-outputs" from the channel decoder to compute the marginal distribution of only the reliable features during likelihood computation at the speech recognizer. This soft-feature error concealment techniques reduces the ASR error rate by up to four times for certain channels. Also considered is a channel decoding technique with source information that improves ASR performance. F

Efficient scalable encoding for distributed speech recognition

“…Transmission over error-prone transport channels will typically result in some packets being lost. To mitigate the effect of these losses, frame concealment techniques, such as insertion, interpolation and regeneration (Milner and Semnani, 2000;Mayorga et al, 2002) and unequal forward error correction techniques (Weerackody et al, 2002;Riskin et al, 2001) can be used. Now it is required that each packet be independently decodable.…”

“…Separate prediction loops are maintained for the base and each of the enhancement layers. The proposed scalable encoding technique offers several advantages: (i) it provides multiple layers, where the enhancement layer data refines the base layer data to provide higher fidelity representation, (ii) the rate of each layer can be easily modified, and (iii) it provides flexibility for using unequal error protection schemes (Weerackody et al, 2002) to ensure adequate protection for the bitstream. This is especially useful in mobile channels where the channel conditions are constantly changing.…”

Automatic Speech Recognition on Mobile Devices and over Communication Networks

“…Convolutional codes also provide added flexibility to adjust the final protection Performance of media-independent FEC codes, with respect to the level of redundancy inserted, can be enhanced by providing for each bit in the bitstream the adequate level of protection against channel errors. The bits comprising the transmitted datastream of DSR systems do not have equal effect on the word recognition performance of the overall system Weerackody et al 2002). For example, in the case of scalar quantization of extracted speech features, the most significant bits (MSBs) generally provide more information for the recognizer than the least significant bits (LSBs).…”

Multiple Description Quantization for Recognizing Voice Over Packet Networks

“…However, transmitting VQ data over packet networks may experience packet loss and consequently leads to degraded recognition performance. There has been much interest in the use of packet-level forward error correction (FEC) to compensate for packet loss, based on parity codes and Reed-Solomon codes [2] [3]. All of the FEC mechanisms send some redundant information which is based on previously transmitted packets.…”

“…By insertion of the optimal reproduction vector as in(2) and, due to the mutual independence of the erasure channels,…”