A signature verification method is based on a comparison of the dynamics of a reference and a sample signature. Acceleration and pressure signals produced by a known person when writing his or her signature are stored and used as a reference signals. Then, at a later time, a person whose signature is to be verified writes his or her signature to produce acceleration and pressure signals that are compared to the reference signals. The process of comparison involves segmenting the two sets of signals to facilitate identifying regions of high probable correlation and then correlating corresponding segment pairs. Segmentation is based on pen lifts which represent reproducible timing marks in the signatures. According to the disclosed method, a pen or other writing instrument is used which produces a signal representative of the first time derivative of the pressure forces exerted on the stylus of the pen. The second time derivative of the pressure forces is computed from the measured signal. The first and second time derivatives are examined to detect quiet times indicative of lifting the pen from a writing surface. The detected quiet times are checked to determine if they have a time duration which exceeds a predetermined time period. The polarity and amplitude of the second time derivatives are tested to determine if they are positive and exceed a predetermined threshold at the beginning and end of each detected quiet time. Preferably, the steps of detecting quiet times, examining the duration of the quiet times, and testing the polarity and amplitude of the second derivatives are performed concurrently.
Coordinates of a handwritten pattern drawn on a tablet are sequentially sampled by a pattern recognition unit to prepare pattern coordinate data. Based on an area encircled by segments created by the sampled pattern coordinate data of one stroke and a line connecting a start point and an end point of the one-stroke coordinate data, the sampled pattern coordinate data of the one stroke is converted to a straight line and/or curved line segments. The converted segments are quantized and normalized. The segments of the normalized input pattern are rearranged so that the input pattern is drawn in a predetermined sequence. Differences between direction angles for the rearranged segments are calculated. Those differences are compared with differences of the direction angles of the dictionary patterns read from a memory to calculate a difference therebetween. The matching of the input pattern and the dictionary pattern is determined in accordance with the difference. If the matching fails, the first or last inputted segment of the input pattern is deleted or the sampled pattern coordinate data of the next stroke is added, to continue the recognition process.
A signature to be verified is written on an area which carries horizontal and vertical lines using a pen. As the lines are crossed signals representative of the vertical and horizontal components of the pen tip velocity and whether the tip is in contact with the area are passed to a computer. A hidden Markov model derived from vertical and horizontal velocities and a "contact" signal occurring as a number of authentic signatures are written is stored by the computer. A forward pass of the Baum-Welch algorithm is used to calculate the probability that the vertical and horizontal components of the pen tip and the contacts between pen tip and the writing area could have been produced from the hidden Markov model. This probability is used to decide whether the signature is authentic. The hidden Markov model stored by the computer is derived from an initial model based on pen tip velocities and contacts occurring in an authentic signature, and re-estimation carried out using forward and backward passes of the Baum-Welch algorithm and velocities and contacts from further authentic signatures.
A method for storing compressed data indicative of an analog signal and reproducing that analog signal from the compressed data. The analog signal to be stored is superimposed on a rectangular coordinate system. An elementary interval is defined on the X-axis of an X-Y coordinate system for the selection of a first series of points, a second series of points (R.sub.1, R.sub.2, R.sub.3) is determined by establishing a threshold ratio between the chord (C.sub.1, C.sub.1 ') separating two non-consecutive points and the sum of the chords (C.sub.1, C.sub.2, C.sub.3 ; C.sub.1 ', C.sub.2 ', C.sub.3 ') separating each of the consecutive points within two non-consecutive points. An interval (N) between these two non-consecutive points is selected to correspond to the maximum ratio below this threshold. The analog signal is reconstituted by generating a signal having a curve that is plotted by establishing between each group of three non-consecutive points a parabola with respect to the axis of symmetry corresponding to the axis of the elementary intervals.
An image identification apparatus for identifying an image from a hand drawn representation of at least part of the image, the image identification apparatus comprising an image processor arranged in operation to generate a reference identification in response to spatial samples produced from at least part of the hand drawn representation, the reference identification being indicative of a first estimate of which of a plurality of pre-stored images corresponds to the hand drawn representation. and a controller which is arranged in operation to cause the image processor to produce a refined reference identification from the spatial samples and further spatial samples produced from a further part of the hand drawn representation, the refined reference identification being indicative of a refined estimate of which of the plurality of the prestored images corresponds to the hand drawn representation. The controller is arranged to operate with the image processor to generate further refined estimates of the identification of the image from subsequent hand drawn parts of the image.
A computer-based system and method for recognizing handwriting. The present invention includes a pre-processor, a front end, and a modeling component. The present invention operates as follows. First, the present invention identifies the lexemes for all characters of interest. Second, the present invention performs a training phase in order to generate a hidden Markov model for each of the lexemes. Third, the present invention performs a decoding phase to recognize handwritten text. Hidden Markov models for lexemes are produced during the training phase. The present invention performs the decoding phase as follows. The present invention receives test characters to be decoded (that is, to be recognized). The present invention generates sequences of feature vectors for the test characters by mapping in chirographic space. For each of the test characters, the present invention computes probabilities that the test character can be generated by the hidden Markov models. The present invention decodes the test character as the recognized character associated with the hidden Markov model having the greatest probability.
