Saburo Tazaki

Asymptotic performance of block quantizers with difference distortion measures

Gershos bounds on the asymptotic (large rate or small distortion) performance of block quantizers are valid for vector distortion measures that are powers of the Euclidean or l_{2} norm. These results are generalized to difference distortion measures that are increasing functions of the seminorm of their argument, where any seminorm is allowed. This provides a k -dimensional generalization of Gish and Pierces results for single-symbol quantizers. When the distortion measore is a power of a seminorm the bounds are shown to be strictly better than the corresponding bounds provided by the k th-order rate-distortion functions.

Improvement on error rate performance for FM recording code

A partial response (PR) equalization technique for an FM recording code is proposed. First, the noise power spectrum and the eye patterns at the discrimination point for the PR system are compared with those for Veillards system. Then, the bit-error rates of bit-by-bit decoding and Viterbi coding for these systems are obtained by computer simulation. The results show that the PR system has improved performance, both for bit-by-bit decoding and for Viterbi decoding, over Veillards system. >

Variance mismatch of vector quantizers (Corresp.)

The asymptotic performance of variance-mismatched vector quantizers is presented. It is demonstrated by both asymptotic analysis and computer simulations that well-designed vector quantizers are inherently more invulnerable to mismatch than are conventional scalar quantizers. A generalized exponential density function is considered as a statistical model of sources. As an example, the asymptotic performance is derived and applied to the memoryless Laplacian source with the squared-error distortion measure.

Error Characteristics of FM Family Codes Using PRML

In this paper, we clarify the properties of thirty FM family codes through analyses of their coding rules. We then investigate combinations of these codes with several PRML techniques by simulating bit error rates. The results show that nine FM family codes yield a better SNR than does the conventional FM code at the bit error rate of 10¿4.

Performance comparisons of multi-level codes for high density MO recording

We discuss the performance of the multi-level codes for high density recording on magneto-optical (MO) disks. The performance of multi-level codes is compared with that of run-length limited (RLL) binary codes using the partial response maximum likelihood (PRML) method. The simulations indicate that the multi-level codes have good performance in high density recording and that the PRML method produces little improvement in the performance of multi-level codes compared with those of binary-level codes.

Error correction using IFMV9 coded overlapped PN sequence for power line SSMA

Proposes an IFMV9 coded OL(overlapped)-PN sequence as a PN sequence that is suitable for DS/SS data transmission with low spreading factor through a power line channel. The IFMV9 coded OL-PN sequence is a data stream coded according to the coding rule of IFMV9, that is one of the channel codes, in every unit period length of a PN sequence. It can provide DC free and correlation properties. This coded sequence also makes it possible to combine maximum-likelihood (ML) decoding techniques with the well-known differential decoding method at the receivers side. The authors investigate the error rate characteristics and the possible maximum number of simultaneous users by computer simulation. They use a model of television of which a power circuit impedance alternates rapidly and greatly, as the simulated channel. The simulated result shows the error rate improvement due to the combination of ML decoding and differential decoding at the receiving end.<<ETX>>

PRML Method on 4/15 Code and 4/11 Code

A 4/15 code and a 4/11 code have been developed for a magnetooptic disk system with a sampling servo mechanism. We propose the application of PR systems instead of differential detection methods. We also adopt Viterbi detection with PR systems as PRML methods. This paper also proposes another signal detection method which makes use of the unique property of the coding rule for the 4/15 code. The bit error rates of our systems were examined by computer simulation and compared with conventional differential detection. The results show that our proposed systems exhibit excellent performance compared with conventional systems.

A proposal on a generalized method for constructing the fixed-length adaptive block recording code

A simple and more generalized algorithm is presented for the design of a fixed-length adaptive block recording code, named the (d, k; m, n) code where d and k are the minimum and maximum run-length of the same symbol, and m and n are the dataword and codeword lengths. This algorithm is based on systematic use of both lookback and lookahead informations and an expanded version of the merging rule by Jacoby. Applications of this algorithm to (2, 9; 3, 6) and (2, 8; 4, 8) NRZI waveform codes are demonstrated.

A Construction Method for Picture Information Having a Recovery Function

There has been a recent trend for keeping important information on social activities by outputting it as picture information on the paper medium, which goes against the current trend of doing away with paper. Among the methods developed to follow this trend, those in which preserving the high quality of the picture image is desired are numerous. On the other hand, picture information processing is conducted in general from the point of view of information engineering in recovering the characters of ancient books, for example. However, there are limitations in the conventional methods since studies on recovering picture information have not been carried out for these methods. If the state of development of the signal processing technology in recent years is considered, conducting processing that supports the recovery of preserved picture in advance is considered to be most effective. In this paper, from the viewpoint of constructing a forward-looking picture information system, conducting processing allowing the recovery function for digital picture information output on paper, and so on is proposed and a scheme for realizing it is discussed. In the proposed scheme, the original picture image information can be reproduced from the picture information preserved on paper with erasures.

BER performance of baseband DS/SSMA using FM coded M sequences in bandwidth limited channels

We discuss the bit error rate (BER) performance of asynchronous baseband DS/SSMA systems using line coded M sequences. As the line codes, we adopt FM family codes that provide M sequences with the DC free spectrum and good correlation properties. It has been already reported that some combinations of three FM coded M sequences provide a better cross-correlation property than that of the preferred pairs of noncoded M sequences. Since their quantitative evaluation in terms of the BER performance improvement due to the line coding has not been done yet, we considered it in this paper. We first derive an equation for the BER analysis, then compare the BER performance of the FM coded M sequences with that of the noncoded M sequences in two bandwidth limited channels, a band pass and a power line channel. In our scenario, the improvement of the required E/sub b//N/sub 0/ is 1 dB and 3 dB, respectively. Furthermore we propose a system using FM coded M sequences of different periods simultaneously. This proposed system of the different periods can achieve the channel efficiency 4/3 times better than a system using noncoded M sequences of the same period, while the required E/sub b//N/sub 0/ of both systems is almost the same.