Abstract: An apparatus for determining an estimated pitch lag is provided. The apparatus includes an input interface for receiving a plurality of original pitch lag values, and a pitch lag estimator for estimating the estimated pitch lag. The pitch lag estimator is configured to estimate the estimated pitch lag depending on a plurality of original pitch lag values and depending on a plurality of information values, wherein for each original pitch lag value of the plurality of original pitch lag values, an information value of the plurality of information values is assigned to the original pitch lag value.

Abstract: Curation of a corpus of a cognitive computing system is performed interactively by reporting on user interface device to a user a parse tree illustration of discrepancies and corresponding assigned confidence factors detected between at least a portion of a first document and a second or more documents in the corpus. Responsive to a user selection of an illustrated discrepancy in the parse tree, a drill-down dialog is prepared and displayed which shows at least a text string for the portion of the first document and at least one conflicting text string from the second or more documents, and which provides at least one user-selectable administrative action option for handling the detected discrepancy. Responsive to receipt of user selection of an administrative action option, the computing system performs the action to handle the detected discrepancy.

Abstract: Systems, methods, and computer-readable storage devices for crowd-sourced data labeling. The system requests a respective response from each of a set of entities. The set of entities includes crowd workers. Next, the system incrementally receives a number of responses from the set of entities until one of an accuracy threshold is reached and m responses are received, wherein the accuracy threshold is based on characteristics of the number of responses. Finally, the system generates an output response based on the number of responses.

Abstract: The present invention extends to methods, systems, and computer program products for detecting an event from signals in a listening area. Listening area formation is triggered based on detecting a potential event from a normalized signal, based on characteristics of a normalized signal, or based on a signal pattern anomaly. A geo cell database is queried with search terms derived from one or more normalized signals. A geo cell subset is returned from the geo cell database. A listening area is formed from the geo cells included in the geo cell subset. A listener monitors an additional normalized signal in the listening area. An event is detected and/or validated (or rejected) based on the additional normalized signal.

Abstract: A device for automatic translation of patent claims automatically translates a first language claim into a second language claim. The device includes an element noun retrieving module, a complement determining module, an organizing module. and an automatic computer translating module. The element noun retrieving module retrieves an element noun from the first language claim. The complement determining module determines a complement definition from the first language claim, as a combination of an interrelationship definition and object element. The organizing module forms an organized structure by organizing the element noun, interrelationship definition, object element and an organization identifying word.

Abstract: Background noise estimators and methods are disclosed for estimating background noise in an audio signal. Some methods include obtaining at least one parameter associated with an audio signal segment, such as a frame or part of a frame, based on a first linear prediction gain, calculated as a quotient between a residual signal from a 0th-order linear prediction and a residual signal from a 2nd-order linear prediction for the audio signal segment. A second linear prediction gain is calculated as a quotient between a residual signal from a 2nd-order linear prediction and a residual signal from a 16th-order linear prediction for the audio signal segment. Whether the audio signal segment comprises a pause is determined based at least on the obtained at least one parameter; and a background noise estimate is updated based on the audio signal segment when the audio signal segment comprises a pause.

Abstract: Speech is transcribed to produce a transcript. At least some of the text in the transcript is encoded as data. These codings may be verified for accuracy and corrected if inaccurate. The resulting transcript is provided to a decision support system to perform functions such as checking for drug-drug, drug-allergy, and drug-procedure interactions, and checking against clinical performance measures (such as recommended treatments). Alerts and other information output by the decision support system are associated with the transcript. The transcript and associated decision support output are provided to a physician to assist the physician in reviewing the transcript and in taking any appropriate action in response to the transcript.

Abstract: Provided is a voice interaction apparatus including an estimation unit configured to estimate whether to wait a next user's speech or to execute an output of an output speech by the voice interaction apparatus based on a user's speech input by an input unit; and the response control unit configured to, when the estimation unit could not estimate whether to wait for the user's next speech or to execute the output of the output speech by the voice interaction apparatus, respond by executing at least one of a filler output and a nodding action, the filler output being an output of a filler, i.e., a connecting utterance irrelevant to an interaction content, from the output unit, and the nodding action being an action of swinging a pseudo head part back and forth.

Abstract: A hybrid speech translation system whereby a wireless-enabled client computing device can, in an offline mode, translate input speech utterances from one language to another locally, and also, in an online mode when there is wireless network connectivity, have a remote computer perform the translation and transmit it back to the client computing device via the wireless network for audible outputting by client computing device. The user of the client computing device can transition between modes or the transition can be automatic based on user preferences or settings. The back-end speech translation server system can adapt the various recognition and translation models used by the client computing device in the offline mode based on analysis of user data over time, to thereby configure the client computing device with scaled-down, yet more efficient and faster, models than the back-end speech translation server system, while still be adapted for the user's domain.

Abstract: A speech interface device may capture user speech for analysis by automatic speech recognition (ASR) and natural language understanding (NLU) components. However, an audio signal representing the user speech may also contain interfering sound generated by a media player that is playing audio content such as music. Before performing ASR and NLU, a system attempts to identify the content being played by the media player, such as by querying the media player or by analyzing the audio signal. The system then obtains the same content from an available source and subtracts the audio represented by the content from the audio signal.

Abstract: An approach to keyword spotting makes use of acoustic parameters that are trained on a keyword spotting task as well as on a second speech recognition task, for example, a large vocabulary continuous speech recognition task. The parameters may be optimized according to a weighted measure that weighs the keyword spotting task more highly than the other task, and that weighs utterances of a keyword more highly than utterances of other speech. In some applications, a keyword spotter configured with the acoustic parameters is used for trigger or wake word detection.

Abstract: A system for generating comfort noise for a stream of frames carrying an audio signal includes frame characterizing logic configured to generate a set of filter parameters characterising the frequency content of a frame; an analysis filter adapted using the filter parameters and configured to filter the frame so as to generate residual samples; an analysis controller configured to cause the residual samples to be stored in a store responsive to receiving an indication that the frame does not comprise speech; and a synthesis controller operable to select stored residual samples from the store and cause a synthesis filter, inverse to the analysis filter and adapted using filter parameters generated by the frame characterizing logic for one or more frames not comprising speech, to filter the selected residual samples so as to generate a frame of comfort noise.

Abstract: Predicting and learning users' intended actions on an electronic device based on free-form speech input. Users' actions can be monitored to develop a list of carrier phrases having one or more actions that correspond to the carrier phrases. A user can speak a command into a device to initiate an action. The spoken command can be parsed and compared to a list of carrier phrases. If the spoken command matches one of the known carrier phrases, the corresponding action(s) can be presented to the user for selection. If the spoken command does not match one of the known carrier phrases, search results (e.g., Internet search results) corresponding to the spoken command can be presented to the user. The actions of the user in response to the presented action(s) and/or the search results can be monitored to update the list of carrier phrases.

Abstract: In one aspect, a first device includes at least one processor, a microphone accessible to the at least one processor, a communication interface accessible to the at least one processor, and storage accessible to the at least one processor. The storage bears instructions executable by the at least one processor to await first voice input comprising a response cue and receive, from the microphone, the first voice input comprising the response cue. The instructions are also executable by the at least one processor to, responsive to receipt of the first voice input comprising the response cue, transmit a wake up command to a second device different from the first device. The wake up command is transmitted via the communication interface using a peripheral device communication protocol.

Abstract: The quality of encoded signals can be improved by reclassifying AUDIO signals carrying non-speech data as VOICE signals when periodicity parameters of the signal satisfy one or more criteria. In some embodiments, only low or medium bit rate signals are considered for re-classification. The periodicity parameters can include any characteristic or set of characteristics indicative of periodicity. For example, the periodicity parameter may include pitch differences between subframes in the audio signal, a normalized pitch correlation for one or more subframes, an average normalized pitch correlation for the audio signal, or combinations thereof. Audio signals which are re-classified as VOICED signals may be encoded in the time-domain, while audio signals that remain classified as AUDIO signals may be encoded in the frequency-domain.

Abstract: Example implementations disclosed herein can be used to generate a local sound signal corresponding to utterances of a user and other sounds detected by a microphone array coupled to a communication device and to condition the local sound signals to separate the utterances of the user from the other sounds to generate a conditioned sound signal. The conditioned sound signals are evaluated to generate a local quality score for the conditioned sound signals, and when the local quality score of the conditioned sound signals is below a threshold associated with the communication device, a local feedback message indicating a local user position change can be generated. The local feedback message can include instructions for the user to move to another location to improve the quality of the condition sound signals.

Abstract: Monitoring and analysis of a user's speech to detect symptoms of a mental health disorder by continuously monitoring a user's speech in real-time to generate audio data based, transcribing the audio data to text and analyzing the text of the audio data to determine a sentiment of the audio data is disclosed. A trained machine learning model may be applied to correlate the text and the determined sentiment to clinical information associated with symptoms of a mental health disorder to determine whether the symptoms are a symptom event. The initial determination may be transmitted to a second device to determine (and/or verify) whether or not the symptom event was falsely recognized. The trained machine learning model may be updated based on a response from the second device.

Abstract: One embodiment provides a method, including: receiving, at an information handling device, voice data; receiving, at the information handling device, user identification data from at least one user device; determining, using a processor, a voice signature subset from a database of users, wherein the voice signature subset comprises voice signatures associated with a plurality of users identified using the user identification data; and identifying at least one user associated with the voice data by comparing the received voice data to the voice signature subset. Other aspects are described and claimed.

Abstract: Audio information is monitored by a user device that performs audio content recognition of any received audio content. The user device includes a scheduling logic unit, a probe, and an audio signature generator. The scheduling logic unit maintains a set of scheduling rules that define conditions that were present when previous audio content recognition of audio content received by the user device was successful. The scheduling logic unit receives currently present conditions of the user device, and compares the currently present conditions to the set of scheduling rules to determine if the currently present conditions match any scheduling rules. The user devices captures ambient audio content via the probe and generates audio signatures of the captured audio content using the audio signature generator if a match occurs, and inhibits capturing audio content by the user device and/or inhibits generating audio signatures if a match does not occur.

Abstract: A system is provided that employs a statistical approach to semi-supervised speech enhancement with a low-order non-negative matrix factorization (“NMF”). The system enhances noisy speech based on multiple dictionaries with dictionary atoms derived from the same clean speech samples and generates an enhanced speech representation of the noisy speech by combining, for each dictionary, a clean speech representation of the noisy speech generated based on a NMF using the dictionary atoms of the dictionary. The system generates frequency-domain (“FD”) clean speech sample representations of the clean speech samples, for example, using a Fourier transform. To generate each dictionary, the system generates a dictionary-unique initialization of the dictionary atoms and the activations and performs a NMF of the FD clean speech samples.