Abstract: A speech recognition system (10) having a sampler block (12) and a feature extractor block (14) for extracting time domain and spectral domain parameters from a spoken input speech into a feature vector. A polynomial expansion block (16) generates polynomial coefficients from the feature vector. A correlator block (20), a sequence vector block (22), an HMM table (24) and a Viterbi block (26) perform the actual speech recognition based on the speech units stored in a speech unit table (18) and the HMM word models stored in the HMM table (24). The HMM word model that produces the highest probability is determined to be the word that was spoken.

Abstract: The step sizes for tap positions of an adaptive filter (124) in an echo canceller (125) for canceling an echo signal are individually selected based upon a profile of the echo environment. A training signal (112) is injected into the echo canceller and to the echo path during a training interval, or period. The adaptive filter profiles the echo path during the training interval. Respective step sizes &agr;[k] for the tap positions are generated as a function of the echo profile so generated.

Abstract: A voice decoder detects channel errors and loss of cryptographic synchronization using the change in spectral energy between sequential frames. The change in energy between frames is determined between corresponding LSP's of said successive frames and summed together. A running average of the change in energy for a predetermined number of frames is maintained. Current voice frames are eliminated based on the difference between the change in energy associated with the current frame and the running average. Accordingly, offensive audio associated with such channel errors or cryptographic synchronization loss is eliminated.

Abstract: A speech recognition system (10) having a sampler block (12) and a feature extractor block (14) for extracting time domain and spectral domain parameters from a spoken input speech into a feature vector. A polynomial expansion block (16) generates polynomial coefficients from the feature vector. A correlator block (20), a sequence vector block (22), an HMM table (24) and a Veterbi block (26) perform the actual speech recognition based on the speech units stored in a speech unit table (18) and the HMM word models stored in the HMM table (24). The HMM word model that produces the highest probability is determined to be the word that was spoken.

Abstract: The step sizes for tap positions of an adaptive filter (124) in an echo canceller (125) for canceling an echo signal are individually selected based upon a profile of the echo environment. A training signal (112) is injected into the echo canceller and to the echo path during a training interval, or period. The adaptive filter profiles the echo path during the training interval. Respective step sizes &agr;[k] for the tap positions are generated as a function of the echo profile so generated.

Abstract: A hands-free communications unit employs a downlink peak power memory element (150, FIG. 1), a downlink voice activity detector (140), and an uplink voice activity detector (180). The downlink voice activity detector (140) evaluates the incoming signal to determine if the signal includes voice. In a similar manner, the uplink voice activity detector (180) determines if an outgoing signal includes voice. In the event that the incoming signal includes voice and the energy of the outgoing signal exceeds a threshold calculated from the peak power of the incoming signal, the uplink voice activity detector (180) indicates that a double-talk condition exists (350, FIG. 3). In the event that the incoming signal does not include voice, the uplink voice activity detector (180) compares the uplink signal to a voice threshold above a noise floor to determine if the outgoing signal includes voice.

Abstract: A hands-free communications unit employs a downlink peak power memory element (150, FIG. 1), a downlink voice activity detector (140), and an uplink voice activity detector (180). The downlink voice activity detector (140) evaluates the incoming signal to determine if the signal includes voice. In a similar manner, the uplink voice activity detector (180) determines if an outgoing signal includes voice. In the event that the incoming signal includes voice and the energy of the outgoing signal exceeds a threshold calculated from the peak power of the incoming signal, the uplink voice activity detector (180) indicates that a double-talk condition exists (350, FIG. 3). In the event that the incoming signal does not include voice, the uplink voice activity detector (180) compares the uplink signal to a voice threshold above a noise floor to determine if the outgoing signal includes voice.

Abstract: A technique and method that efficiently and simultaneously transfers both voice and data information over a single communications link (11). The technique and method utilizes a multi-rate vocoder (12) for providing voice information at a plurality of different rates. Accordingly, the vocoder may be dynamically adjusted during transfer of information to meet channel bandwidth requirements. The technique and method also utilizes a bit-control protocol for inserting voice information within a stream of data information by the use of inserting flags having predetermined bitstreams. Accordingly, voice information is dynamically inserted into an existing bitstream of data information such that voice information is transferred only upon demand and when needed. This prevents the transfer of useless information, such as voice information when a speaker is silent and, thus, allows for very efficient utilization of system bandwidth.

Abstract: Apparatus and a method are described for providing a security token. The security token may be used for numerous applications, but is particularly useful in conjunction with radiotelephone devices having secure functions. In accordance with the invention the user must insert a PIN and utter sounds. The radiotelephone device will activate the secure functions only if the PIN is correct and the sounds uttered are authenticated against voice feature vectors stores it the token.

Abstract: A method and system simplifies and accomplishes the programming of a PED (12), such as a cellular telephone (22), via the use of a personal computer (PC). The present invention includes software (52, 54) within the PC (14) for providing a graphical user interface (GUI) to the user for ease and simplicity of selecting various programming features and settings associated with the cellular telephone to be programmed. The software further includes a translator (56) for translating these user selected features/settings into specific key depresses to be sent to the cellular telephone, via software and hardware interfaces (58, 18, 20), for accomplishing the programming of such features/settings. Additionally, the software includes the capability to monitor the data being displayed on the cellular telephone because such data may be required to accomplish various programming features.

Abstract: A two-wire line and a four-wire line telephone system interface with each other through a hybrid (12). After a connection between the two and four-wire line systems is made, an echo canceler (10) provides training signals to the hybrid on the transmit path an generates echo cancellation signals on the receive path. An adaptive filter (16) is employed to provide the proper echo cancellation signals, while a controller (22) provides filter coefficients based on an error signal. The training signals include the convolution of an impulse function with a sin(x)/x function. Once the filter coefficients are determined for the particular call, the transmit path of the four-wire line system is switched in and voice data between the two systems my commence.

Abstract: An echo canceller (500) utilizes a central control unit (502) (CCU) and at least one convolution processing unit (504) (CPU) to perform echo cancellation on one or more channels. A method for operating the echo canceller (500) controls reading (706) one or more reference signals (510) and reflected signals (512), convolution and adaptive processing (11, 21, 31, 41) using the CPUs (504), and post-processing (702) to produce one or more echo cancelled signals (514). Convolution and adaptive processing for multiple channels can be performed simultaneously (11, 31) by multiple CPUs (504), or in a time-multiplexed manner (11, 21) by a single CPU (504), or a combination (11, 21, 31, 41) of simultaneous and time-multiplexing by multiple CPUs (504). CCU (502) is able to perform processing functions (702-730) while CPUs (504) perform convolution and adaptive processing (11, 21, 31, 41).

Abstract: A quadrant detector circuit (400) has a comparator (442) having a pair of inputs (438, 440). A first (438) of the pair of inputs (438, 440) is coupled to an in-phase signal (434) and a second (440) of the pair of inputs (438, 440) is coupled to a quadrature phase signal (436). A sample counter (448) has a reset (446) coupled to an output (444) of the comparator (442). A controllable switch (456) has a selection input (454) coupled to an output (452) of the sample counter (448). The controllable switch (456) is capable of switching between a local oscillator signal (458) and an inverse local oscillator signal (460).

Abstract: A balanced differential radio receiver (100) has an antenna (64) operative for receiving a modulated signal. A first pair of mixers (110, 112) each having an input that receives an output (106, 108) of the antenna (64) and each of the first pair of mixers (110, 112) having an output (118, 120). A second pair of mixers (114, 116) each having an input that receives an inverse output (108, 106) of the antenna (64). A resistive adding circuit (122, 124) having a pair of inputs (118, 120). The inputs (118, 120) coupled to the output of the first pair of mixers (110, 112).

Abstract: A technique and method that efficiently and simultaneously transfers both voice and data information over a single communications link (11). The technique and method utilizes a multi-rate vocoder (12) for providing voice information at a plurality of different rates. Accordingly, the vocoder may be dynamically adjusted during transfer of information to meet channel bandwidth requirements. The technique and method also utilizes a byte-control protocol for inserting voice information within a stream of data information by the use of a data link escape (DLE) byte command (34). Accordingly, voice information is dynamically inserted into an existing stream of data information such that voice information is transferred only upon demand and when needed. This prevents the transfer of useless information, such as voice information when a speaker is silent and, thus, allows for very efficient utilization of system bandwidth.

Abstract: The present invention includes an adaptive filter (40) for use within an echo canceler. The adaptive filter operates in a first mode for performing Decimation Echo Location (DEL). During the DEL mode, high energy regions occurring within the echo impulse response are identified over an extended time range through signal decimation. The adaptive filter then operates in a second mode for performing Partitioned Echo Cancellation (PEC). During the PEC mode, filter coefficients not associated with these high energy regions are modeled as zero while the identified high energy region filter coefficients of the impulse response are partitioned and convolution of the reference signal, x(n), with the impulse response signal, h(n), is accomplished by performing convolution on only those samples in the reference signal memory block (50) that correspond to respective high energy region filter coefficients in the adaptive filter block (56).