We consider the joint sequence estimation, timing and phase recovery for linear modulation. The paper differs from the classical ones in the sense that time-discrete algorithms suitable for fully digital receivers are discussed. Sufficient conditions are given such that the signal samples represent sufficient statistics. These conditions involve signal bandwidth, sampling/symbol rate and the analog prefilter characteristics. It is shown that the sampling rate need not be an exact multiple of the symbol rate, i.e., the samples can be taken from a free-running oscillator. All subsequent signal processing operations in the receiver then operate with the clock of this free-running oscillator. Timing recovery is then performed by a time-variant linear digital interpolator and a decimator. Carrier recovery and sequence estimation are performed at an average rate of one symbol per sample. The digital matched filter for this case is derived for an arbitrary colored noise spectrum.>

This paper gives a status report of the Philips research system for phoneme-based, large-vocabulary, continuous-speech recognition. Like for many other systems, the recognition architecture is based on an integrated statistical approach. We describe the characteristic features of the system as opposed to other systems: 1. The Viterbi criterion is consistently applied both in training and testing. 2. Continuous mixture densities are used without tying or smoothing. 3. Time-synchronous beam search in connection with a phoneme look-ahead is applied to a tree-organized lexicon. The system has been successfully applied to the American English DARPA RM task. Here, we report experimental results for a German 13 000-word Philips internal dictation task. In addition to the scientific prototype, a PC version has been set up which is described here for the first time.

The author describes the improvements in a time synchronous beam search strategy for a 10000-word continuous speech recognition task. The improvements are based on two measures: a tree-organization of the pronunciation lexicon and a novel look-ahead technique at the phoneme level, both of which interact directly with the detailed search at the state levels of the phoneme models. Experimental tests were performed for four speakers on a 12306-word task. As a result of the above measures, the overall search effort was reduced by a factor of 17 without a loss in recognition accuracy.>

Using the maximum-likelihood approach, algorithms for detection and synchronization are derived that are well suited for VLSI implementation. Special emphasis is placed on an all-digital implementation where carrier and clock synchronization do not require a feedback of signals to the analog part, which simplifies the analog front-end design (mixing oscillator and A/D converter sampling clock run at fixed frequency). An important advantage of the proposed algorithms is that a high clock rate is not required; only two-four times the symbol rate is needed, depending on amplitude quantization. Implementation aspects, e.g. architecture, and quantization, are considered. A prototype is described which was implemented to prove the feasibility of the concept and to evaluate the performance under practical conditions.>

The digital realization of timing recovery circuits for digital data transmission is considered. A digital algorithm is proposed that can be implemented very efficiently even at high data rates. The resulting timing jitter has been computed and verified by simulations. In contrast to other known algorithms, the one presented here allows free-running sampling oscillators and a novel planar filtering method that prevents synchronization hangups.>

In a recent paper [1], Gardner described a new timing-error detector for sampled receivers. The detector derives timing information without decision direction from a BPSK or QPSK signal sampled at rate2/T(symbol intervalT). We will show how this detector can be derived from the maximum likelihood principle, and thus give insight into the nature of the algorithm.