Radio telecommunication device and method of authenticating a user with a voice authentication token

Claim 1

A radio frequency (RF) portable telecommunication device, comprising: a token receiving portion adapted to receive and read an encrypted personal identification number and encrypted first voice feature vectors of spoken words stored in a voice authentication token when said voice authentication token is inserted into the token receiving portion; a keypad for permitting a user to enter an identification number; apparatus for decrypting the encrypted personal identification number; a first comparator for comparing said decrypted personal identification number with said entered personal identification number; voice receiving circuitry for receiving predetermined sounds comprised of speech segments from said user and for extracting second voice feature vectors from said speech segments; and apparatus for decrypting said encrypted first voice feature vectors; a second comparator for comparing said second voice feature vectors with said decrypted first voice feature vectors, wherein predetermined secure functions of said telecommunication device are enabled when said first and said second comparators both indicate that a match has occurred, said secure features being enabled only while said authentication token is present in said token receiving portion, wherein the voice receiving circuitry comprises: a converter for converting the predetermined sounds to digital speech signals; an end point detector for detecting a beginning and an end of each of the speech segments; a first and second low-pass filter for filtering each detected speech segment of the digital speech signals to produce digitally filtered voice signals; energy management circuitry for producing an energy estimate for each detected speech segment based on an averaged normalized gain for each detected speech segment; a linear predictive analyzer for performing a linear predictive analysis on the digitally filtered voice signals produced by the second low-pass filter to determine a set of linear predictive coefficients for each detected speech segment; a pitch detector for performing a pitch detection on the linear predictive coefficients to produce a pitch lag and pitch gain for each detected speech segment; a formant analyzer for performing a formant analysis on the digitally filtered voice signals to determine formant frequency locations for each speech segment; a cepstral analyzer for performing a cepstral analysis on the digitally filtered voice signals to produce cepstral high-time and low-time estimates for each speech segment; and a dynamic time-warping element for forming the second voice feature vectors by performing a dynamic time-warping function on said energy estimate, the set of linear predictive coefficients, the pitch lag and gain, the formant frequency locations and the cepstral high-time and low-time estimates.

Claim 8

A method of authenticating a user of a portable RF telecommunications device, wherein first voice feature vectors of spoken sounds of the user and a personal identification number are stored in an encrypted form on a voice authentication token, the method comprising the steps of: receiving said voice authentication token placed in said telecommunications device; receiving an entered personal identification from a proposed user; decrypting the encrypted personal identification number stored on the voice authentication token; comparing the entered personal identification number with the decrypted personal identification number; reading said first voice feature vectors from said voice authentication token; prompting said proposed user to speak predetermined sounds comprised of speech segments; forming second voice feature vectors from said speech segments; comparing said first and second voice feature vectors; enabling secure features of said telecommunications device when said first and second voice feature vectors match, wherein the reading, prompting, forming and comparing the first and second voice feature vector steps are performed when the entered personal identification number matches the decrypted personal identification number, and wherein the forming second voice feature vectors step comprises the steps of: converting the predetermined sounds to digital speech signals; applying the digital speech signals to an end point detector to detect a beginning and an end of each of the speech segments; filtering each detected speech segment of the digital speech signals with a first and second low-pass filter to produce digitally filtered voice signals; applying the digitally filtered voice signals produced by the first low-pass filter to energy management circuitry, the energy management circuitry producing an energy estimate for each detected speech segment based on an averaged normalized gain for each detected speech segment; performing a linear predictive analysis on the digitally filtered voice signals produced by the second low-pass filter to determine a set of linear predictive coefficients for each detected speech segment; performing a pitch detection on the linear predictive coefficients to produce a pitch lag and pitch gain for each detected speech segment; performing a formant analysis on the digitally filtered voice signals to determine formant frequency locations for each speech segment; performing a cepstral analysis on the digitally filtered voice signals to produce cepstral high-time and low-time estimates for each speech segment; and forming the second voice feature vectors by performing a dynamic time-warping function on said energy estimate, the set of linear predictive coefficients, the pitch lag and gain, the formant frequency locations and the cepstral high-time and low-time estimates.

Claim 9

A method of authenticating a user of a portable RF telecommunications device, wherein first voice feature vectors of spoken sounds of the user and a personal identification number are stored in an encrypted form on a voice authentication token, the method comprising the steps of: receiving said voice authentication token placed in said telecommunications device; receiving an entered personal identification from a proposed user; decrypting the encrypted personal identification number stored on the voice authentication token; comparing the entered personal identification number with the decrypted personal identification number; reading said first voice feature vectors from said voice authentication token; prompting said proposed user to speak predetermined sounds comprised of speech segments; forming second voice feature vectors from said speech segments; comparing said first and second voice feature vectors; enabling secure features of said telecommunications device when said first and second voice feature vectors match, wherein the reading, prompting, forming and comparing the first and second voice feature vector steps are performed when the entered personal identification number matches the decrypted personal identification number, and wherein the forming second voice feature vectors step comprises the steps of: converting the predetermined sounds to digital speech signals; applying the digital speech signals to an end point detector to detect a beginning and an end of each of the speech segments; filtering each detected speech segment of the digital speech signals with a first and second low-pass filter to produce digitally filtered voice signals; applying the digitally filtered voice signals produced by the first low-pass filter to energy management circuitry, the energy management circuitry producing an energy estimate for each detected speech segment based on an averaged normalized gain for each detected speech segment; performing a linear predictive analysis on the digitally filtered voice signals produced by the second low-pass filter to determine a set of linear predictive coefficients for each detected speech segment; performing a pitch detection on the linear predictive coefficients to produce a pitch lag and pitch gain for each detected speech segment; performing a formant analysis on the digitally filtered voice signals to determine formant frequency locations for each speech segment; performing a cepstral analysis on the digitally filtered voice signals to produce cepstral high-time and low-time estimates for each speech segment; and forming the second voice feature vectors by performing a dynamic time-warping function on said energy estimate, the set of linear predictive coefficients, the pitch lag and gain, the formant frequency locations and the cepstral high-time and low-time estimates.