Claims Analysis
Total Claims: 13
Claim 1
Independent Claim
222 words
A method of verifying a finite signal pattern with various features, each subject to quantitative variability, involving the following steps; a. in a learning phase, supplying a predetermined number of signal patterns and for each characteristic drawing up a histogram giving the number of times that the characteristic concerned had a value within a range of values of a collection of adjacent value ranges applicable to the feature; b. in a test phase, supplying the signal pattern to be verified and detecting the value for each of said characteristics in order to indicate the value range belonging to said feature within the collection appropriate to said feature; c. multiplying the histogram values of the value ranges thus indicated for all said features of the signal to be verified to give an end product wherein every factor of said end product is different from zero; wherein said step further comprises: d. in order to standarize a signal pattern supplied, multiplying the detected value of each feature by a factor (1+e) where e<<1 to give a corrected value, and repeating this each time for all corrected values thus found until an aggregate value derived from the corrected values applying exceeds an aggregate threshold; e. comparing said end product with a standard value in order to signal verification only if the standard value is exceeded.
Claim 2
Independent Claim
222 words
A method of verifying a finite signal pattern with various features, each subject to quantitative variability, involving the following steps; a. in a learning phase, supplying a predetermined number of signal patterns and for each characteristic drawing up a histogram giving the number of times that the characteristic concerned had a value within a range of values of a collection of adjacent value ranges applicable to the feature; b. in a test phase, supplying the signal pattern to be verified and detecting the value for each of said characteristics in order to indicate the value range belonging to said feature within the collection appropriate to said feature; c. multiplying the histogram values of the value ranges thus indicated for all said features of the signal to be verified to give an end product wherein every factor of said end product is different from zero; wherein said step further comprises: d. in order to standarize a signal pattern supplied, multiplying the detected value of each feature by a factor (1+e) where e<<1 to give a corrected value, and repeating this each time for all corrected values thus found until an aggregate value derived from the corrected values applying exceeds an aggregate threshold; e. comparing said end product with a standard value in order to signal verification only if the standard value is exceeded.
Claim 3
Dependent Claim
23 words
A method as claimed in claim 1, further including: deriving the total signal value by the summation of the values of the features.
Claim 4
Dependent Claim
29 words
A method as claimed in claim 1 further including, after the learning phase, forming the similarity threshold from the multiplication of the maximum occurrency values of the individual features.
Claim 5
Dependent Claim
48 words
A method as claimed in claim 1 further including, after a successful verification, a fresh learning phase is carried out in which the occurrency values of those partial ranges into which the values of the features of the signal to be verified have fallen are increased by 1.
Claim 6
Dependent Claim
26 words
A method as claimed in claim 4, further including dividing all occurrency values by a constant divisor after an occurrency value has exceeded a predetermined value.
Claim 7
Dependent Claim
63 words
A method as claimed in claim 1 wherein the signal is a voice signal of limited duration and the features are adjacent frequency spectrum ranges, whereby the energy of the voice signal is integrated separately in successive equal time intervals per spectrum range and the individual integrals are separately summated over the entire voice signal and the spectrum sums represent the feature values.
Claim 8
Dependent Claim
92 words
A method as claimed in claim 6, wherein, in order to establish the partial ranges into which the individual spectrum sums fall, during the voice signal the partial sum obtained for each spectrum range after every summation of a further integral is compared with the upper limit of a partial range and the content of an intermediate store location allocated to the spectrum range concerned is raised by 1 whenever the partial sum exceeds the limit, whereupon the comparison with the upper limit of the next higher partial range is carried out.
Claim 9
Independent Claim
41 words
A method as claimed in one of claims 6, wherein the frequencies obtained during the sampling phase are stored on a data support and automatically read in and stored to verify on the supply of the voice signal to be verified.
Claim 10
Dependent Claim
227 words
An arrangement for implementing the method as claimed in claim 1 comprising, a feature store (51) with a store location for each partial range of each feature; a first addressing arrangement (53) which addresses the store locations of all partial ranges of a feature and a second addressing arrangement (29, 41) which addresses a partial range every time; a range limit store (23) controlled by at least the second addressing arrangement; a comparator device (25) which compares the values of the features with a corresponding range limit read out from the range limit store (23) and produces a signal at a value smaller than the compared range limit but greater than the range limit of the preceding range limit; said signal reading out the store address of the feature store (51) addressed at the same time as the range limit store and during the sampling phase raising the read out content in an occurrency value adder (55) by 1 and writing it again into the same store location and during the test phase takes it to a processing unit (91, 93, 95) which multiplies this content by the contents of the store locations read out for the other features and takes it to a similarity comparator (97), the other input of which receives the similarity threshold and the output (98) of which supplies the result of verification.
Claim 11
Dependent Claim
263 words
An arrangement as claimed in claim 9 for the verification of a limited-duration voice signal, whereby the features are adjacent frequency spectrum ranges, wherein the energy of the voice signal is separately integrated in consecutive equal time intervals per spectrum range and the individual integrals are separately summated over the entire voice signal and the spectrum values represent the values of the features, whereby the electrical voice signal generated by a microphone passes through a filter bank which divides the voice signal into a number of frequency spectrum ranges and separately integrates the energy in the individual spectrum ranges during consecutive equal-length time intervals and the integrated signals are taken to an accumulator from spectrum value adder and a long-term spectrum store with a store location allocated to each spectrum range, which summates the integrated signals per spectrum range, whereby the comparator arrangement (25) receives the intermediate sum of each spectrum range generated by the spectrum value adder (19) and compares it with the range limit read out from the range limit store (23), in which the second addressing arrangement includes an intermediate store (29) with a special store location per spectrum range which is addressed by the first addressing arrangement (53) and from which the content read out is increased via a further adder (27) by one unit if the comparator arrangement (25) shows that the intermediate sum is greater than the range limit read out, said content thereafter being re-written into the intermediate store (29), and where the output of the intermediate store provides the addresses of the second addressing arrangement.
Claim 12
Dependent Claim
40 words
An arrangement as claimed in claim 10, wherein the range limit store (29) receives a special range limit value for each feature and each partial range and where the first addressing arrangement (53) also addresses the range limit store (29).
Claim 13
Dependent Claim
28 words
An arrangement as claimed in claim 10, wherein the feature store (51) contains the value 1 in all store addresses before the start of the first sampling signal.