Amazon Technologies, Inc.

Audible, Inc.

Thomas Schaaf

Stan Weidner Salvador

Leah Rose Nicolich-Henkin

Mingzhi Yu

Hamza Lakhani

Ricardo DeMatos

Sumedha Arvind Kshirsagar

Jeffrey C O'Neill

Jeffrey Paul Lilly

Remus Razvan Mois

Rafal Kuklinski

Derek Christopher Murman

Darya Chernikhova

Roger Alix-Gaudreau

Techniques for confirming an operator of a vehicle is drowsy are described. A vehicle computing system sends data (e.g., raw sensor data and/or alert data corresponding to an indication that a driver is impaired determined based on the raw sensor data) to a remote server(s). The remote server(s) confirms the driver is impaired based on the raw sensor data and/or other contextual data. The remote server(s) then receives output data from a speechlet and causes the vehicle computing system to present output audio corresponding to output data.

Vehicle voice user interface

In some cases, a handheld device that includes a microphone and a scanner may be used for voice-assisted scanning. For example, a user may provide a voice input via the microphone and may activate the scanner to scan an item identifier (e.g., a barcode). The handheld device may communicate voice data and item identifier information to a remote system for voice-assisted scanning. The remote system may perform automatic speech recognition (ASR) operations on the voice data and may perform item identification operations based on the scanned identifier. Natural language understanding (NLU) processing may be improved by combining ASR information with item information obtained based on the scanned identifier. An action may be executed based on the likely user intent.

Voice-assisted scanning

Techniques for engaging a drowsy or otherwise impaired driver of a vehicle in a VUI dialog are described. A vehicle computing system sends data (e.g., raw sensor data and/or an indication that a driver is impaired determined based on the raw sensor data) to a remote server(s). The remote server(s) may separately determine whether the driver is impaired based on the raw sensor data and/or other contextual data. The remote server(s) selects a speechlet to provide output data based on the sensor data, contextual data, and or a level at which the driver is impaired. The remote server(s) then causes the vehicle computing system to present output audio corresponding to output data provided by the speechlet.

Systems and methods are provided for automated analysis of audio recordings and/or automated recommendations of audio recordings. The audio recordings may be automatically analyzed based on power, frequency, tempo, complexity, dimensionality, other audio features, and/or some combination thereof. Automated recommendations of audio recordings may be provided based at least on the variability levels. User interfaces may be provided to receive user feedback of preferred audio recordings and to select audio recordings to recommend based on the user feedback and the automated analysis. The automated recommendations may be further based on user purchase data and/or listening history.

Automated recommendations of audio narrations

An audio buffer is used to capture audio in anticipation of a user command to do so. Sensors and processor activity may be monitored, looking for indicia suggesting that the user command may be forthcoming. Upon detecting such indicia, a circular buffer is activated. Audio correction may be applied to the audio stored in the circular buffer. After receiving the user command instructing the device to process or record audio, at least a portion of the audio that was stored in the buffer before the command is combined with audio received after the command. The combined audio may then be processed, transmitted or stored.

Smart circular audio buffer

Features are disclosed for recognizing inappropriate content in an output. The offensive content may be generated as a result of a speech processing error. A system may identify the inappropriate elements of a generated output and select among different appropriate alternatives. The system may be adjusted based on certain user characteristics. The system may be localized based on language and cultural features. The system may modify the generated output based on characteristics such as the tolerance threshold of known persons in the proximity of the system. The tolerance threshold may further be used to personalize and modify available content. Models used by the system may be further trained using input from a user.

Controlling offensive content in output

Features are provided for selectively scoring portions of user utterances based at least on articulatory features of the portions. One or more articulatory features of a portion of a user utterance can be determined. Acoustic models or subsets of individual acoustic model components (e.g., Gaussians or Gaussian mixture models) can be selected based on the articulatory features of the portion. The portion can then be scored using a selected acoustic model or subset of acoustic model components. The process may be repeated for the multiple portions of the utterance, and speech recognition results can be generated from the scored portions.

Growing community of inventors

Pittsburgh, PA, United States of America

Thomas Schaaf