Abstract: In order to improve the accuracy of recognition of hand-written input using a stylus, output signals from a plurality of accelerometers representing x and y axis acceleration and deceleration are sampled at a predetermined rate and passed through a digital bandpass filter to remove high frequency components and dc components arising from gravity. X and y vectors derived from the original x and y input signals are passed to a classifier using a hidden Markov model. Bandpass filtering improves the robustness of the interpretation of the vectors against stored templates which may be templates of individual characterizations or of whole words.

Abstract: A speech recognizer suitable for distributed speech recognition is robust to missing speech feature vectors. Speech is transmitted via a packet switched network in the form of basic feature vectors. Missing feature vectors are detected and replacement feature vectors are estimated by interpolation of received data prior to speech recognition. Features may be converted and interpolation may be accomplished in a spectral domain.

Abstract: Calculate the log frame energy value of each of a pre-determined number n of frames of an input speech signal and apply a matrix transform to the n log frame energy values to form a temporal matrix representing the input speech signal. The matrix transform may be a discrete cosine transform.