Abstract: Described herein are systems and methods for improved audio analysis using a computer-executed neural network having one or more in-network data augmentation layers. The systems described herein help ease or avoid unwanted strain on computing resources by employing the data augmentation techniques within the layers of the neural network. The in-network data augmentation layers will produce various types of simulated audio data when the computer applies the neural network on an inputted audio signal during a training phase, enrollment phase, and/or testing phase. Subsequent layers of the neural network (e.g., convolutional layer, pooling layer, data augmentation layer) ingest the simulated audio data and the inputted audio signal and perform various operations.

Abstract: Utterances of at least two speakers in a speech signal may be distinguished and the associated speaker identified by use of diarization together with automatic speech recognition of identifying words and phrases commonly in the speech signal. The diarization process clusters turns of the conversation while recognized special form phrases and entity names identify the speakers. A trained probabilistic model deduces which entity name(s) correspond to the clusters.

Abstract: A system for generating channel-compensated features of a speech signal includes a channel noise simulator that degrades the speech signal, a feed forward convolutional neural network (CNN) that generates channel-compensated features of the degraded speech signal, and a loss function that computes a difference between the channel-compensated features and handcrafted features for the same raw speech signal. Each loss result may be used to update connection weights of the CNN until a predetermined threshold loss is satisfied, and the CNN may be used as a front-end for a deep neural network (DNN) for speaker recognition/verification. The DNN may include convolutional layers, a bottleneck features layer, multiple fully-connected layers and an output layer. The bottleneck features may be used to update connection weights of the convolutional layers, and dropout may be applied to the convolutional layers.

Abstract: A method of obtaining and automatically providing secure authentication information includes registering a client device over a data line, storing information and a changeable value for authentication in subsequent telephone-only transactions. In the subsequent transactions, a telephone call placed from the client device to an interactive voice response server is intercepted and modified to include dialing of a delay and at least a passcode, the passcode being based on the unique information and the changeable value, where the changeable value is updated for every call session. The interactive voice response server forwards the passcode and a client device identifier to an authentication function, which compares the received passcode to plural passcodes generated based on information and iterations of a value stored in correspondence with the client device identifier. Authentication is confirmed when a generated passcode matches the passcode from the client device.

Abstract: Embodiments described herein provide for automatically classifying the types of devices that place calls to a call center. A call center system can detect whether an incoming call originated from voice assistant device using trained classification models received from a call analysis service. Embodiments described herein provide for methods and systems in which a computer executes machine learning algorithms that programmatically train (or otherwise generate) global or tailored classification models based on the various types of features of an audio signal and call data. A classification model is deployed to one or more call centers, where the model is used by call center computers executing classification processes for determining whether incoming telephone calls originated from a voice assistant device, such as Amazon Alexa® and Google Home®, or another type of device (e.g., cellular/mobile phone, landline phone, VoIP).

Abstract: Disclosed are systems and methods including computing-processes executing machine-learning architectures for voice biometrics, in which the machine-learning architecture implements one or more language compensation functions. Embodiments include an embedding extraction engine (sometimes referred to as an “embedding extractor”) that extracts speaker embeddings and determines a speaker similarity score for determine or verifying the likelihood that speakers in different audio signals are the same speaker. The machine-learning architecture further includes a multi-class language classifier that determines a language likelihood score that indicates the likelihood that a particular audio signal includes a spoken language. The features and functions of the machine-learning architecture described herein may implement the various language compensation techniques to provide more accurate speaker recognition results, regardless of the language spoken by the speaker.

Abstract: Embodiments described herein provide for performing a risk assessment using graph-derived features of a user interaction. A computer receives interaction information and infers information from the interaction based on information provided to the computer by a communication channel used in transmitting the interaction information. The computer may determine a claimed identity of the user associated with the user interaction. The computer may extract features from the inferred identity and claimed identity. The computer generates a graph representing the structural relationship between the communication channels and claimed identities associated with the inferred identity and claimed identity. The computer may extract additional features from the inferred identity and claimed identity using the graph. The computer may apply the features to a machine learning model to generate a risk score indicating the probability of a fraudulent interaction associated with the user interaction.

Abstract: Disclosed are systems and methods including computing-processes executing machine-learning architectures implementing label distribution loss functions to improve age estimation performance and generalization. The machine-learning architecture includes a front-end neural network architecture defining a speaker embedding extraction engine of the machine-learning architecture, and a backend neural network architecture defining an age estimation engine of the machine-learning architecture. The embedding extractor is trained to extract low-level acoustic features of a speaker's speech, such as mel-frequency cepstral coefficients (MFCCs), from audio signals, and then extract a feature vector or speaker embedding vector that mathematically represents the low-level features of the speaker. The age estimator is trained to generate an estimated age for the speaker and a Gaussian probability distribution around the estimated age, by applying the various types of layers of the age estimator on the speaker embedding.

Abstract: The present invention is directed to a deep neural network (DNN) having a triplet network architecture, which is suitable to perform speaker recognition. In particular, the DNN includes three feed-forward neural networks, which are trained according to a batch process utilizing a cohort set of negative training samples. After each batch of training samples is processed, the DNN may be trained according to a loss function, e.g., utilizing a cosine measure of similarity between respective samples, along with positive and negative margins, to provide a robust representation of voiceprints.

Abstract: Embodiments described herein provide for passive caller verification and/or passive fraud risk assessments for calls to customer call centers. Systems and methods may be used in real time as a call is coming into a call center. An analytics server of an analytics service looks at the purported Caller ID of the call, as well as the unaltered carrier metadata, which the analytics server then uses to generate or retrieve one or more probability scores using one or more lookup tables and/or a machine-learning model. A probability score indicates the likelihood that information derived using the Caller ID information has occurred or should occur given the carrier metadata received with the inbound call. The one or more probability scores be used to generate a risk score for the current call that indicates the probability of the call being valid (e.g., originated from a verified caller or calling device, non-fraudulent).

Abstract: A computer may train a single-class machine learning using normal speech recordings. The machine learning model or any other model may estimate the normal range of parameters of a physical speech production model based on the normal speech recordings. For example, the computer may use a source-filter model of speech production, where voiced speech is represented by a pulse train and unvoiced speech by a random noise and a combination of the pulse train and the random noise is passed through an auto-regressive filter that emulates the human vocal tract. The computer leverages the fact that intentional modification of human voice introduces errors to source-filter model or any other physical model of speech production. The computer may identify anomalies in the physical model to generate a voice modification score for an audio signal. The voice modification score may indicate a degree of abnormality of human voice in the audio signal.

Abstract: Aspects of the invention determining a threat score of a call traversing a telecommunications network by leveraging the signaling used to originate, propagate and terminate the call. Outer-edge data utilized to originate the call may be analyzed against historical, or third party real-time data to determine the propensity of calls originating from those facilities to be categorized as a threat. Storing the outer edge data before the call is sent over the communications network permits such data to be preserved and not subjected to manipulations during traversal of the communications network. This allows identification of threat attempts based on the outer edge data from origination facilities, thereby allowing isolation of a compromised network facility that may or may not be known to be compromised by its respective network owner.

Abstract: Embodiments include a computer executing voice biometric machine-learning for speaker recognition. The machine-learning architecture includes embedding extractors that extract embeddings for enrollment or for verifying inbound speakers, and embedding convertors that convert enrollment voiceprints from a first type of embedding to a second type of embedding. The embedding convertor maps the feature vector space of the first type of embedding to the feature vector space of the second type of embedding. The embedding convertor takes as input enrollment embeddings of the first type of embedding and generates as output converted enrolled embeddings that are aggregated into a converted enrolled voiceprint of the second type of embedding.

Abstract: Disclosed are systems and methods including computing-processes executing machine-learning architectures extract vectors representing disparate types of data and output predicted identities of users accessing computing services, without express identity assertions, and across multiple computing services, analyzing data from multiple modalities, for various user devices, and agnostic to architectures hosting the disparate computing service. The system invokes the identification operations of the machine-learning architecture, which extracts biometric embeddings from biometric data and context embeddings representing all or most of the types of metadata features analyzed by the system. The context embeddings help identify a subset of potentially matching identities of possible users, which limits the number of biometric-prints the system compares against an inbound biometric embedding for authentication.

Abstract: Disclosed are systems and methods including computing-processes executing machine-learning architectures extract vectors representing disparate types of data and output predicted identities of users accessing computing services, without express identity assertions, and across multiple computing services, analyzing data from multiple modalities, for various user devices, and agnostic to architectures hosting the disparate computing service. The system invokes the identification operations of the machine-learning architecture, which extracts biometric embeddings from biometric data and context embeddings representing all or most of the types of metadata features analyzed by the system. The context embeddings help identify a subset of potentially matching identities of possible users, which limits the number of biometric-prints the system compares against an inbound biometric embedding for authentication.

Abstract: A computer may train a single-class machine learning using normal speech recordings. The machine learning model or any other model may estimate the normal range of parameters of a physical speech production model based on the normal speech recordings. For example, the computer may use a source-filter model of speech production, where voiced speech is represented by a pulse train and unvoiced speech by a random noise and a combination of the pulse train and the random noise is passed through an auto-regressive filter that emulates the human vocal tract. The computer leverages the fact that intentional modification of human voice introduces errors to source-filter model or any other physical model of speech production. The computer may identify anomalies in the physical model to generate a voice modification score for an audio signal. The voice modification score may indicate a degree of abnormality of human voice in the audio signal.

Abstract: An automated speaker verification (ASV) system incorporates a first deep neural network to extract deep acoustic features, such as deep CQCC features, from a received voice sample. The deep acoustic features are processed by a second deep neural network that classifies the deep acoustic features according to a determined likelihood of including a spoofing condition. A binary classifier then classifies the voice sample as being genuine or spoofed.

Abstract: Systems, methods, and computer-readable media for call classification and for training a model for call classification, an example method comprising: receiving DTMF information from a plurality of calls; determining, for each of the calls, a feature vector including statistics based on DTMF information such as DTMF residual signal comprising channel noise and additive noise; training a model for classification; comparing a new call feature vector to the model; predicting a device type and geographic location based on the comparison of the new call feature vector to the model; classifying the call as spoofed or genuine; and authenticating a call or altering an IVR call flow.

Abstract: The present invention is directed to a deep neural network (DNN) having a triplet network architecture, which is suitable to perform speaker recognition. In particular, the DNN includes three feed-forward neural networks, which are trained according to a batch process utilizing a cohort set of negative training samples. After each batch of training samples is processed, the DNN may be trained according to a loss function, e.g., utilizing a cosine measure of similarity between respective samples, along with positive and negative margins, to provide a robust representation of voiceprints.

Abstract: Embodiments described herein provide for passive caller verification and/or passive fraud risk assessments for calls to customer call centers. Systems and methods may be used in real time as a call is coming into a call center. An analytics server of an analytics service looks at the purported Caller ID of the call, as well as the unaltered carrier metadata, which the analytics server then uses to generate or retrieve one or more probability scores using one or more lookup tables and/or a machine-learning model. A probability score indicates the likelihood that information derived using the Caller ID information has occurred or should occur given the carrier metadata received with the inbound call. The one or more probability scores be used to generate a risk score for the current call that indicates the probability of the call being valid (e.g., originated from a verified caller or calling device, non-fraudulent).