Abstract: A maximum likelihood decoding system includes a branch metric processor which calculates only one of four branch metrics associated with branches leading to two consecutive states S.sub.j and S.sub.+1, where j is even. The system determines the remaining three metrics by producing a second branch metric by manipulating the first branch metric using simple binary operations and assigns the first and second metrics to the second and first branches, respectively, leading to the odd state. The system next retrieves associated path metrics from a path metric memory which stores the information in locations accessed by addresses related to identifiers associated with the branch initial states. After the system selects a surviving path for each end state, it stores in a path memory location associated with the end state information identifying the previous state on the surviving path.

Abstract: A noise-suppression circuit (10) divides the signal from a microphone (12) into a plurality of frequency sub-bands by means of a noise-band divider (18) and a subtraction circuit (36). By means of gain circuits (32) and (34), it applies separate gains to the separate bands and then recombines them in a signal combiner (38) to generate an output signal in which the noise has been suppressed. Separate gains are applied only to the lower subbands in the voice spectrum. Accordingly, the noise-band divider (18) is required to compute spectral components for only those bands. By employing a sliding-discrete-Fourier-transform method, the noise-band divider (18) computes the spectral components on a sample-by-sample basis, and circuitry (50, 52) for determining the individual gains can therefore update them on a sample-by-sample basis, too.

Abstract: The channel-impulse-response estimator (44') in a maximum-likelihood-sequence-estimation receiver produces channel-model parameters for a model that produces outputs not only for symbol times but also for intermediate times between the symbol times. A symbol-sequence-derivation circuit (42') determines the most-likely sequence on the basis of metrics computed from the differences between the received signal and the responses of the model to candidate sequences not only at the symbol times but also at the intermediate times. An interpolator (100) receives the symbol decisions from the symbol-sequence-derivation circuit (42') and generates intermediate values from them by simulating the Nyquist filter formed by the concatenation of pulse-shaping filters (22 and 24) in the transmitter and matched filters (36 and 37) in the receiver. The channel-impulse-response estimator (44') uses these values together with the derived-symbol values as inputs to its updating process to maintain the model.