Abstract: A reduced-complexity, fast converging adaptive filter may be used for network echo cancellation applications, including applications having sparse echo paths. The new filter, referred to as a selective-partial-update proportionate NLMS filter, may be based on a proportionate NLMS (PNLMS) technique and selective partial updating of the adaptive filter coefficients. The new PNLMS filter may exploit sparseness of a communications channel to speed up the initial convergence of the NLMS technique included in the filter by weighting regressor data proportionately with an estimated magnitude of the channel impulse response. Selective partial updating is essentially a data selection method to reduce the computational complexity. The performance of the selective-partial-update PNLMS filter compares favorably to an adaptive filter using standard PNLMS for echo paths specified in ITU-T Recommendation G.168.

Abstract: Detectors determine the presence of valid sinusoids for DTMF, MF-R1 and MF-R2 protocols for encoding dialed digits. The detectors split electrical signals into subbands. Energies within the subbands are analyzed to determine a presence of sinusoids corresponding to frequencies of dialed digits. In one embodiment, the detectors comprise a PS-IIR filter to split the electrical signal into the subbands. The detectors further comprise at least one bank of filters, such as notch filters, corresponding to the number of possible relevant frequencies within the respective subbands. The detectors further comprise detection logic comprising tests, which may include analyzing the output(s) from the bank of filters. Optionally, a preclassifier is employed to predetermine which filters in the banks of filters are to be executed. The detectors, typically deployed in digital signal processors, allow for an increase in the density of detectors and provide robust performance in talk-off situations.

Abstract: A signal classifier decomposes a subject signal into subbands. The signal classifier then analyzes the subbands for a presence of energy and uses detection logic to classify the energy as being representative of a frame of a signal using one of the following protocols: DTMF, MF-R1, ANS (V.25), LEC-DIS, V.21, or AA (i.e., telephone, facsimile, or modem). The detection logic is expandable to detect future protocols. The classifier may employ a preclassifier to select a reduced number of notch filters from among plural notch filters at frequencies which sinusoidal signals composing the energy in the subbands of the protocols may be found. A look-up table having predetermined ranges of data corresponding to the frequencies of the sinusoidal signals may be used to reduce calculation times. By splitting the subject signal into subbands and using efficient filters, the classifier uses an order of magnitude fewer processor instruction cycles than previous classifiers.