Method and arrangement for controlling user access
Method for controlling user access to a service available in a data network and/or to information stored in a user database, in order to protect stored user data from unauthorized access, such that the method comprises the following: input of a user's speech sample to a user data terminal, processing of the user's speech sample in order to obtain a prepared speech sample as well as a current voice profile of the user, comparison of the current voice profile with an initial voice profile stored in an authorization database, and output of an access-control signal to either permit or refuse access, taking into account the result of the comparison step, such that the comparison step includes a quantitative similarity evaluation of the current and the stored voice profiles as well as a threshold-value discrimination of a similarity measure thereby derived, and an access-control signal that initiates permission of access is generated only if a prespecified similarity measure is not exceeded.
1. Method for controlling user access to a service made available in a data network and/or to information stored in a user database, in order to protect stored user data from unauthorized access, the method comprises:
inputting a speech sample of a user to a user data terminal,
processing the user's speech sample in order to obtain a prepared speech sample as well as a current voice profile of the user,
and outputting an access-control signal to permit or refuse access, taking into account the result of the comparison step, wherein the comparison step includes: comparing an input speech sample with a speech sample model to determine a first similarity measure, and comparing the first similarity measure with an upper threshold value, wherein access is denied if the upper threshold value is exceeded; and if the upper threshold value is not exceeded, comparing, using a voice-profile comparator unit, the current voice profile with an initial voice profile that has been stored in an authorization database to obtain a second similarity measure, and then comparing the second similarity measure with a lower threshold value, and an access control signal that initiates permission of access is generated only if the second similarity measure exceeds the lower threshold value.
2. Method according to claim 1 , wherein the upper threshold value of the similarity measure is determined dynamically on the basis of several stored voice profiles of at least one of a user, many users or several currently input speech samples.
3. Method according to claim 1 , wherein the speaker is required to provide multiple, identical current speech samples, and each of the speech samples so provided is used for the first similarity measure and thereafter or previously at least one of the current speech samples is used to calculate the current voice profile and the latter, with inclusion of the second similarity measure, is compared with the stored stimulus profile or one of them, wherein a result of the comparison of the first similarity measure determines whether the comparison with the second similarity measure is to be performed.
4. Method according to claim 1 , wherein both the initial and every additional voice profile obtained during a user's attempt to gain access is stored.
5. Method according to claim 4 , wherein either continually or during a current attempt to gain access a current upper threshold of the similarity measure is determined from the stored voice profiles.
6. Method according to claim 4 , wherein the current voice profile is compared with more than one of the stored voice profiles, from the results of this comparison a mean current similarity measure is determined, and a threshold discrimination is applied to the mean current similarity measure.
7. Method according to claim 1 , wherein during the initialization and subsequent access attempts, speech samples provided by the user are stored, and a group of initial similarity evaluations is applied to the currently input speech sample, and thereafter or previously the comparison of the second similarity measure is applied to the voice profile derived from the currently input speech sample in comparison with the stored voice profile or one of them.
8. Method according to claim 1 , wherein a user-guidance output instructs the user to input a first speech sample to which he was previously trained, as well as a second, not previously trained speech sample, and a combined similarity evaluation is applied to the first and second speech samples.
9. Method according to claim 1 , wherein as similarity measure a measure, of a distance separating the current and a stored voice profile is calculated, such that this procedure can in particular include a likelihood consideration.
10. Arrangement for implementing a user access control to a data network and/or to information stored in a user database to protect stored user data from unauthorized access, comprising:
speech-sample input means for the input of speech samples,
speech-sample processing means for to obtaining a prepared speech sample,
speech-sample storage means for storing prepared speech samples,
voice-profile calculation means for obtaining a voice profile from an input speech sample,
voice-profile storage means for storing the voice profiles thus obtained,
a voice-profile comparator unit to compare various voice profiles,
an access-control unit connected on an input side with an output of the voice-profile comparator unit, to emit an access-control signal, such that a similarity-measure calculation unit associated with the voice-profile-comparator unit on its output side is connected to a first and a second threshold-value discriminator for discriminating a calculated similarity measure at a lower and an upper threshold value, such that an input speech sample is compared with a speech sample model to determine a first similarity measure, and the first similarity measure is compared with the upper threshold value in the first threshold value discriminator, and access is denied if the upper threshold value is exceeded, and if the upper threshold value is not exceeded, the voice-profile comparator unit compares the current voice profile with an initial voice profile that has been stored in an authorization database to obtain a second similarity measure, and the second similarity measure is compared in the second threshold value discriminator with the lower threshold value, the output of at least the second threshold-value discriminator is connected regarding the control signal to a control input of the access-control unit.
11. Arrangement according to claim 10 , wherein at least one of the first or second threshold-value discriminator is designed to be dynamically adjustable.
12. Arrangement according to claim 10 , wherein connected to the output of the speech-sample processing means is a speech-sample comparator unit, with which is associated a second similarity-measure calculation unit which on its output side is connected to the second threshold-value discriminator, wherein the similarity-measure calculation unit of the voice-profile comparator unit acts as the first similarity-measure calculation unit and on its output side is connected to the first threshold-value discriminator.
13. Arrangement according to claim 10 , wherein a user-guidance output unit is provided to emit prespecified speech-sample input requirements, and the voice profile calculation means and the voice-profile comparator unit are designed with multiple channels or for sequential, independent processing of several currently input speech samples for the determination of several current similalrity measures, and wherein the similarity-measure calculation unit, which is likewise designed with multiple channels or operates sequentially, is associated at its output side with a similarity-measure comparator unit which compares several of the similarity measures in the individual channels or several sequentially calculated similarity measures with one another, and wherein the access-control signal is generated while taking into account the result of such comparisons.