Abstract: An apparatus and a method for acoustic scene classification of a block of audio samples are provided. The block is partitioned into frames in the time domain. For each respective frame of a plurality of frames of the block, a change measure between the respective frame and a preceding frame of the block is calculated. The respective frame is assigned, based on the calculated change measure, to one of a set of short-event frames, a set of long-event frames, and a set of background frames. The feature vector is determined based on a feature computed from one or more of the set of short-event frames, the set of long-event frames, and the set of background frames.

Abstract: Disclosed herein are systems and methods for handling unwanted telephone calls. In one aspect, an exemplary method comprises, intercepting a call request for a call from a terminal device of a calling party to a terminal device of a called party, generating a call recording containing media data transmitted within a connection established by the intercepted call request, determining attributes of the generated call recording, classifying the call as an unwanted call based on the determined attributes, wherein the classification is performed by a classifier trained on previously collected unwanted calls, and wherein the call is classified as unwanted when the attributes belong to an unwanted call class that is known, and handling the call in accordance with the classification of the call, the handling including at least securing information of the call.

Abstract: This disclosure describes techniques for establishing a voice authentication credential for an authenticated user of a mobile device. In one example, a system comprises an IVR system and a computing system, where the computing system comprises processing circuitry configured to: receive, over a network, authentication data from a mobile device; authenticate, based on the authentication data, a user operating the mobile device; output, over the network and to the mobile device, instructions for communicating with the IVR system; after outputting the instructions, receive, from the IVR system, information indicating that the mobile device, operated by the user, has contacted the IVR system; communicate with the IVR system to authenticate the user of the mobile device, receive, from the IVR system, a voiceprint created by the user of the mobile device; store the voiceprint as an authentication credential for the user of the mobile device.

Abstract: A computer-implemented method can include obtaining activity data corresponding to an environment where a communication device is located. The activity data can include audio data. The communication device can be configured to transmit the audio data. The method can further include identifying, based at least in part on the activity data, a potential audio disruption. The method can further include determining, based at least in part on the audio data, an activity status of a user of the communication device. The method can further include determining that the activity status is an inactive status. The method can further include modifying, in response to both the identifying the potential audio disruption and the determining that the activity status is the inactive status, a transmission of the audio data from the communication device.

Abstract: A method can include obtaining audio data corresponding to a user of a communication device. The communication device can be configured to transmit the audio data. The method can further include obtaining proximity data indicating a user distance between the user and the communication device. The method can further include determining that the user distance exceeds a threshold distance. The method can further include determining, based at least in part on the audio data, an activity status of the user. The method can further include determining that the activity status is an inactive status. The method can further include modifying, in response to the determining that the threshold distance is exceeded and that the activity status is the inactive status, a transmission of the audio data from the communication device.

Abstract: Systems and methods for analyzing structured data are described. A device may receive a table of structured data and create a Raw Pair Distance (RPD) table. The device then selects a set of nodes from the elements in the RPD table and outputs a nodes table. The device may also output a node-node distance (NND) matrix using the RPD table and run an energy reduction algorithm on the NND matrix in order to create an NSPACE matrix including n-dimensional coordinates for each node. The device may display (e.g., via a 3D visualizer) a graphical representation of selected nodes and coordinated relationships between the selected nodes. The systems and methods may enable a user to quickly search and understand relationships within a large structured data set.

Abstract: A method to adjust volume may include obtaining an audio signal during a communication session between a first device and a second device. The method may also include adjusting a volume level of the audio signal so that the volume level is a particular volume level above a noise floor of the audio signal and directing the adjusted audio signal to a remote transcription system.

Abstract: An echo cancellation method based on delay estimation is provided. In the method, a microphone signal and a reference signal are received and preprocessed. In the preprocessed microphone signal and the preprocessed reference signal, frequency point signals with non-linearity in a current echo cancellation scenario are determined. A current delay estimation value is calculated based on frequency point signals without non-linearity in the microphone signal and the reference signal. The reference signal is shifted based on the current delay estimation value. An adaptive filter is updated based on the preprocessed microphone signal and the shifted reference signal, to perform echo cancellation.

Abstract: An audio processing apparatus includes a linear processing stage and control signal circuitry (CSC). The linear processing stage is configured to receive an input audio signal, and to linearly process the input audio signal based on a user-provided setting of the linear processing stage, so as to produce a linear stage output audio signal. The control signal circuitry (CSC), which is configured to (a) generate a control signal based on (i) the user-provided setting of the linear stage and (ii) a prespecified signal quality preserving criterion, and (b) control the linear stage output audio signal with the control signal, so as to produce a controlled audio signal, in compliance with the signal quality preserving criterion.

Abstract: A method for detecting of fraudulent calls is provided. The method includes initiating, by an origin node, a call to a target node, and generating, by the origin node, a first call data record (CDR) for the call, wherein the CDR includes at least one call parameter of the call recorded by the origin node upon termination of the call. The method also includes generating, by the origin node, a first secure CDR that contains the first CDR, sending, by the origin node, the first secure CDR to the target node, and generating, by the origin node, a fraud notification based on a second secure CDR received from a first node and the first secure CDR.

Abstract: The disclosure describes soft limiting electrical signals generated by a transducer before amplification, buffering or other signal conditioning in a microphone assembly, circuits and methods therefor. The soft limiting circuit includes a first circuit portion that limits a first portion of the electrical signal with reduced distortion when the first portion of the signal exceeds a threshold. A second circuit portion of the soft limiting circuit similarly limits a second portion of the electrical signal. In this way, overloading of an amplifier is avoided without excessive distortion.

Abstract: Systems and methods are disclosed for online data-linked telecommunications decisioning and distribution. One method includes receiving call data relating to a telephone call from a telephone device of a user to an interactive voice response (“IVR”) system; accessing a database storing correlated call data and user data; retrieving correlated call data and user data based on the telephone number of the call data; determining a confidence score defining a confidence that the received call data relates to the retrieved correlated call data and user data; correlating the received call data with retrieved call data and user data when the confidence score is greater than a threshold value; determining an IVR response to present to the user via the IVR system; and transmitting the determined IVR response to the IVR system for presentation to the telephone device of the user.

Abstract: The disclosed teleconferencing methods involve detecting a howl state during a teleconference which involves two or more teleconference client locations and a teleconference server. The teleconference server is configured for providing full-duplex audio connectivity between the teleconference client locations. The howl state is a state of acoustic feedback involving two or more teleconference devices in a teleconference client location. Detecting the howl state involves an analysis of both spectral and temporal characteristics of teleconference audio data. The disclosed teleconferencing methods involve determining which client location is causing the howl state and involve mitigating the howl state or sending a howl state detection message.

Abstract: The invention concerns a mobile terminal with at least two transducers (LSm, LSs1, LSS2) used simultaneously as loudspeakers for stereophonic effect. According to the invention, one of said transducers is a main transducer (LSm) with a main working frequency band (Bm) corresponding to at least the phone frequency band, while the other transducer is a secondary transducer (LSs1; LSs2) with a secondary working frequency band (Bs1; Bs2) band different from said main frequency band, the lowest frequencies of said secondary working frequency band (Bs1; Bs2) being greater than the lowest frequencies of said main working frequency band.

Abstract: Aspects of the technology described herein can provide assisted-communication with an intelligent personal assistant. An exemplary computing device may use a profile handler to receive a user profile of a user and a system profile of an interactive communications system. Moreover, a context handler in the exemplary computing device may receive an indication of a communication event related to the user, such as a call or message, and context information associated with the event. Further, a communication handler in the exemplary computing device may use the context information, the user profile, and/or the system profile to navigate the interactive system for the user. In some instances, where a user is unavailable to address a communication event, the intelligent personal assistant may utilize the communication handler to intercept the communication and negotiate the session on behalf of the user.

Abstract: Systems and methods include a plurality of audio input components configured to generate a plurality of audio input signals, and a logic device configured to receive the plurality of audio input signals, determine whether the plurality of audio signals comprise target audio associated with an audio source, estimate a relative location of the audio source with respect to the plurality of audio input components based on the plurality of audio signals and a determination of whether the plurality of audio signals comprise the target audio, and process the plurality of audio signals to generate an audio output signal by enhancing the target audio based on the estimated relative location. The logic device is further configured to use relative transfer-based covariance to construct directional covariance matrix aligned across frequency bands and find a direction that minimizes beam power subject to distortionless criteria.

Abstract: Parameters of clusters of a context-encoded model are updated, per frequency estimation of an aspect of calls to contacts with respect to unique combinations of contextual data of calls matching the respective clusters. The clusters are weighted according to relevance of the unique combinations of contextual data to current contextual information. First probabilities of calling individual contacts are determined according to the current contextual information and one or more inferences between calls determined from the clusters as weighted. Second probabilities of calling each of the plurality of contacts are determined according to a set of rules defining prioritizations for contacts based on external factors. Third probabilities of calling each of the plurality of contacts are determined according to calling patterns of other of the plurality of contacts. The first, second, and third sets of weights are combined to determine final predictions for predicted probabilities to call each of the contacts.

Abstract: An intelligent transcription and automated record generation system and method are provided in which a record identification module can identify, based on call data from a video call, a record in a customer relationship management (CRM) database that is relevant to the video call. An image recognition and video extraction module can process a video stream that includes video information from the video call to extract relevant portions of the video information that are relevant to the record as extracted video information. A record updater module can automatically modify the record at the CRM database to include at least part of the extracted video information as part of the record.

Abstract: A computer-implemented method for determining potentially undesirable voices, according to some embodiments, includes: receiving a plurality of audio recordings, the plurality of audio recordings comprising voices associated with undesirable activity, and determining a plurality of audio components of each of the plurality of audio recordings. The method may further comprise generating a multi-dimensional vector of audio components, from the plurality of audio components, for each of the plurality of audio recordings to generate a plurality of multi-dimensional vectors of audio components, and comparing audio components between the plurality of multi-dimensional vectors of audio components to determine a plurality of clusters of multi-dimensional vectors, each cluster of the plurality of clusters comprising two or more of the plurality of multi-dimensional vectors of audio components, wherein each cluster of the plurality of clusters corresponds to a blacklisted voice.

Abstract: This disclosure describes techniques for authenticating a person that seeks to engage in a voice call session with a call center agent, an interactive voice response system, or other system. In one example, this disclosure describes a method that includes storing information associating an authorized device with an account; receiving, over a network and from a device operated by a user, authentication credentials for the user; determining, based on the authentication credentials, that the user is authorized to access the account; receiving, over the network and from the device operated by the user, a request to engage in a voice conversation; responsive to receiving the request, accessing the phone number associated with the authorized device; and initiating a voice call session by placing a call, over the network, to the phone number associated with the authorized device.