Techniques regarding speech characterization are provided. For example, one or more embodiments described herein can comprise a system, which can comprise a memory that can store computer executable components. The system can also comprise a processor, operably coupled to the memory, and that can execute the computer executable components stored in the memory. The computer executable components can comprise a speech analysis component that can determine a condition of an origin of an audio signal based on a difference between a first feature of the audio signal and a second feature of a synthesized reference audio signal.

A cough detection system and method uses a first database of physiological information relating to a user for whom cough detection is to be implemented and relating to other people likely to be in the vicinity of the user. A second database (used in real time or as a part of a system calibration) has cough data associated with the physiological information. There is a set of cough detection algorithms, each one tailored to a particular set of physiological characteristics. A cough detection algorithm is selected or constructed which is suitable for identifying coughs of the user while ignoring coughs of the other people. This selected algorithm is applied to sound collected to identify coughs of the user.

A cough detection system and method uses a first database of physiological information relating to a user for whom cough detection is to be implemented and relating to other people likely to be in the vicinity of the user. A second database (used in real time or as a part of a system calibration) has cough data associated with the physiological information. There is a set of cough detection algorithms, each one tailored to a particular set of physiological characteristics. A cough detection algorithm is selected or constructed which is suitable for identifying coughs of the user while ignoring coughs of the other people. This selected algorithm is applied to sound collected to identify coughs of the user.

Systems, devices, and methods for tracking articulatory and prosodic development in children are disclosed. Human speech in a given language can be divided into phonemes, which are a sound or group of sounds perceived by speakers of the language to have a common linguistic function (e.g., consonant sounds, vowel sounds). In an exemplary aspect, a normative model can be generated for production characteristics of each phoneme in a given language using a database of normative speech samples. One or more speech samples of a human subject can be analyzed to identify the phonemes used by the human subject and measured against the normative model. Based on this analysis, a normed score is generated of the articulation accuracy, duration, rhythm, volume, and/or other production characteristics for each phoneme of the speech sample of the human subject.

Trained machine learning models can be used for analysis of signals obtained through an in-ear or on-body device. Signals can be analyzed to determine information related to activities such as eating, chewing, drinking, coughing, or sneezing. In addition, data from an in-ear thermometer or other data sensors can be analyzed in conjunction with the machine learning models to provide data or recommendations to a user on a user device or initiate an action.

Trained machine learning models can be used for analysis of signals obtained through an in-ear or on-body device. Signals can be analyzed to determine information related to activities such as eating, chewing, drinking, coughing, or sneezing. In addition, data from an in-ear thermometer or other data sensors can be analyzed in conjunction with the machine learning models to provide data or recommendations to a user on a user device or initiate an action.

This document describes a data processing system for processing a speech signal for voice-based profiling. The data processing system segments the speech signal into a plurality of segments, with each segment representing a portion of the speech signal. For each segment, the data processing system generates a feature vector comprising data indicative of one or more features of the portion of the speech signal represented by that segment and determines whether the feature vector comprises data indicative of one or more features with a threshold amount of confidence. For each of a subset of the generated feature vectors, the system processes data in that feature vector to generate a prediction of a value of a profile parameter and transmits an output responsive to machine executable code that generates a visual representation of the prediction of the value of the profile parameter.

This document describes a data processing system for processing a speech signal for voice-based profiling. The data processing system segments the speech signal into a plurality of segments, with each segment representing a portion of the speech signal. For each segment, the data processing system generates a feature vector comprising data indicative of one or more features of the portion of the speech signal represented by that segment and determines whether the feature vector comprises data indicative of one or more features with a threshold amount of confidence. For each of a subset of the generated feature vectors, the system processes data in that feature vector to generate a prediction of a value of a profile parameter and transmits an output responsive to machine executable code that generates a visual representation of the prediction of the value of the profile parameter.

At least one speech model, which includes one or more acoustic states exhibited in one or more reference speech samples and defines allowed transitions between the acoustic states, is obtained. At least one test speech sample produced by a subject is received. A plurality of test-sample feature vectors that quantify acoustic features of different respective portions of the test speech sample are computed. The test speech sample is mapped to a minimum-distance sequence of the acoustic states, by mapping the test-sample feature vectors to respective ones of the acoustic states such that a first total distance between the test-sample feature vectors and the respective ones of the acoustic states is minimized. A different, second total distance between the test-sample feature vectors and the respective ones of the acoustic states is computed. Responsively to the second total distance, an output indicating a physiological state of the subject is generated.

At least one speech model, which includes one or more acoustic states exhibited in one or more reference speech samples and defines allowed transitions between the acoustic states, is obtained. At least one test speech sample produced by a subject is received. A plurality of test-sample feature vectors that quantify acoustic features of different respective portions of the test speech sample are computed. The test speech sample is mapped to a minimum-distance sequence of the acoustic states, by mapping the test-sample feature vectors to respective ones of the acoustic states such that a first total distance between the test-sample feature vectors and the respective ones of the acoustic states is minimized. A different, second total distance between the test-sample feature vectors and the respective ones of the acoustic states is computed. Responsively to the second total distance, an output indicating a physiological state of the subject is generated.