Abstract: Disclosed herein are systems, methods, and computer-readable media for classifying a set of inputs via a supervised classifier model that utilizes a novel activation function that provides the capability to learn a scale parameter in addition to a bias parameter and other weight parameters.

Abstract: A system, method and computer program product for classification of an analog electrical signal using statistical models of training data. A technique is described to quantize the analog electrical signal in a manner which maximizes the compression of the signal while simultaneously minimizing the diminution in the ability to classify the compressed signal. These goals are achieved by utilizing a quantizer designed to minimize the loss in a power of the log-likelihood ratio. A further technique is described to enhance the quantization process by optimally allocating a number of bits for each dimension of the quantized feature vector subject to a maximum number of bits available across all dimensions.

Abstract: A method and apparatus for remote access to a target application is disclosed where a system administrator may establish telephonic contact with an interactive voice response system and obtain access to the target application by speech communication. The interactive response system may authenticate the system administrator by implementing various measures including biometric measures. Once access is granted, the interactive response system may broker a communication between the target application using text/data and the system administrator using natural language.

Abstract: A system, method and computer program product for performing blind change detection audio segmentation that combines hypothesized boundaries from several segmentation algorithms to achieve the final segmentation of the audio stream. Automatic segmentation of the audio streams according to the system and method of the invention may be used for many applications like speech recognition, speaker recognition, audio data mining, online audio indexing, and information retrieval systems, where the actual boundaries of the audio segments are required.

Abstract: A method is presented which reduces data flow and thereby increases processing capacity while preserving a high level of accuracy in a distributed speech processing environment for speaker detection. The method and system of the present invention includes filtering out data based on a target speaker specific subset of labels using data filters. The method preserves accuracy and passes only a fraction of the data by optimizing target specific performance measures. Therefore, a high level of speaker recognition accuracy is maintained while utilizing existing processing capabilities.

Abstract: A system, method and computer program product for classification of an analog electrical signal using statistical models of training data. A technique is described to quantize the analog electrical signal in a manner which maximizes the compression of the signal while simultaneously minimizing the diminution in the ability to classify the compressed signal. These goals are achieved by utilizing a quantizer designed to minimize the loss in a power of the log-likelihood ratio. A further technique is described to enhance the quantization process by optimally allocating a number of bits for each dimension of the quantized feature vector subject to a maximum number of bits available across all dimensions.

Abstract: A system, method and computer program product for classification of an analog electrical signal using statistical models of training data. A technique is described to quantize the analog electrical signal in a manner which maximizes the compression of the signal while simultaneously minimizing the diminution in the ability to classify the compressed signal. These goals are achieved by utilizing a quantizer designed to minimize the loss in a power of the log-likelihood ratio. A further technique is described to enhance the quantization process by optimally allocating a number of bits for each dimension of the quantized feature vector subject to a maximum number of bits available across all dimensions.

Abstract: A system and method for verifying user identity, in accordance with the present invention, includes a conversational system for receiving inputs from a user and transforming the inputs into formal commands. A behavior verifier is coupled to the conversational system for extracting features from the inputs. The features include behavior patterns of the user. The behavior verifier is adapted to compare the input behavior to a behavior model to determine if the user is authorized to interact with the system.

Abstract: A technique for improved score calculation and normalization in a framework of recognition with phonetically structured speaker models. The technique involves determining, for each frame and each level of phonetic detail of a target speaker model, a non-interpolated likelihood value, and then resolving the at least one likelihood value to obtain a likelihood score.

Abstract: In automatic pattern recognition, in the context of patterns being observed either in the same or a new environment, e.g. a new acoustic channel, as compared to the one seen during the previous enrollment, an improvement wherein degradation of the system recognition accuracy caused by environment/channel mismatches is averted.

Abstract: A method is presented which reduces data flow and thereby increases processing capacity while preserving a high level of accuracy in a distributed speech processing environment for speaker detection. The method and system of the present invention includes filtering out data based on a target speaker specific subset of labels using data filters. The method preserves accuracy and passes only a fraction of the data by optimizing target specific performance measures. Therefore, a high level of speaker recognition accuracy is maintained while utilizing existing processing capabilities.

Abstract: Methods and arrangements for facilitating data clustering. From a set of input data, a predetermined number of non-overlapping subsets are created. The input data is split recursively to create the subsets.

Abstract: A system, method and computer program product for performing blind change detection audio segmentation that combines hypothesized boundaries from several segmentation algorithms to achieve the final segmentation of the audio stream. Automatic segmentation of the audio streams according to the system and method of the invention may be used for many applications like speech recognition, speaker recognition, audio data mining, online audio indexing, and information retrieval systems, where the actual boundaries of the audio segments are required.

Abstract: Methods and apparatus for the rapid adaptation of classification systems using small amounts of adaptation data. Improvements in classification accuracy are attainable when conditions similar to those that present in adaptation are observed. The attendant methods and apparatus are suitable for a wide variety of different classification schemes, including, e.g., speaker identification and speaker verification.

Abstract: The present invention describes a framework for the analytical and visual analysis and tuning of complex speaker verification systems guided by a policy finite state machine. The Receiver Operating Curve associated with the acoustic speaker recognition task is transformed into a multi-dimensional Receiver Operating Map (ROM), which results from a probabilistic analysis of the policy state machine. A Detection Cost Function (DCF) Map can be similarly generated. Results indicating that optimization over this surface (or the ROM) is an appropriate way to set thresholds are given.

Abstract: A method is presented which reduces data flow and thereby increases processing capacity while preserving a high level of accuracy in a distributed speech processing environment for speaker detection. The method and system of the present invention includes filtering out data based on a target speaker specific subset of labels using data filters. The method preserves accuracy and passes only a fraction of the data by optimizing target specific performance measures. Therefore, a high level of speaker recognition accuracy is maintained while utilizing existing processing capabilities.

Abstract: A technique for improved score calculation and normalization in a framework of recognition with phonetically structured speaker models. The technique involves determining, for each frame and each level of phonetic detail of a target speaker model, a non-interpolated likelihood value, and then resolving the at least one likelihood value to obtain a likelihood score.

Abstract: A system and method for determining and authenticating a person's identity by generating a behavioral profile for that person by presenting that person with various stimulus and measuring that person's response characteristics in an enrollment stage. That person's response profile, once generated is stored. When that user subsequently needs to access a secure resource, that user to be authorized is presented with the stimulus that was presented at the time of generating that person's behavioral profile and the person's responses are detected and compared to his/her behavioral profile. If a match is detected, that user is identified. The user's behavioral response may be in the form of signals as detected by sensor means that detects visual or audible emotional cues or as signals resulting from that person's behavior as detected by polygraph or EEG devices.

Abstract: Automated decision making techniques are provided. For example, a technique for generating a decision associated with an individual or an entity includes the following steps. First, two or more data streams associated with the individual or the entity are captured. Then, at least one time-varying measure is computed in accordance with the two or more data streams. Lastly, a decision is computed based on the at least one time-varying measure. One form of the time-varying measure may include a measure of the coverage of a model associated with previously-obtained training data by at least a portion of the captured data. Another form of the time-varying measure may include a measure of the stability of at least a portion of the captured data. While either measure may be employed alone to compute a decision, preferably both the coverage and stability measures are employed. The technique may be used to authenticate a speaker.

Abstract: Techniques are provided for generating data based on at least one of physical and behavioral characteristics associated with an individual, an entity or a context, which may then be used for a variety of security and/or meta-data analysis purposes. This may be accomplished, for example, by capturing and processing multiple streams of data, such as conversational meta-data, associated with an individual or an entity, and building statistical models and/or extracting heuristics from such data. The statistical models and/or heuristics may then be used, by way of example, for such purposes as user identification, verification of an identity claim, context detection and further heuristic extraction.