Nuance Communications, Inc.

Christophe Ris

Markus Buck

Rudi Vuerinckx

Tim Haulick

Tobias Herbig

Timo Matheja

Martin Roessler

Simon Graf

Jan Vlietinck

Josef Anastasiadis

Stephan Kanthak

Julien Premont

Stijn Verschaeren

Hakan Ari

Andreas Kirbach

Dieter Ranz

There is provided a speech dialog system that includes a first microphone, a second microphone, a processor and a memory. The first microphone captures first audio from a first spatial zone, and produces a first audio signal. The second microphone captures second audio from a second spatial zone, and produces a second audio signal. The processor receives the first audio signal and the second audio signal, and the memory contains instructions that control the processor to perform operations of a speech enhancement module, an automatic speech recognition module, and a speech dialog module that performs a zone-dedicated speech dialog.

Multi-microphone speech dialog system for multiple spatial zones

Methods and apparatus to process microphone signals by a speech enhancement module to generate an audio stream signal including first and second metadata for use by a speech recognition module. In an embodiment, speech recognition is performed using endpointing information including transitioning from a silence state to a maybe speech state, in which data is buffered, based on the first metadata and transitioning to a speech state, in which speech recognition is performed, based upon the second metadata.

Methods and apparatus for speech segmentation using multiple metadata

Deep Neural Networks (DNNs) with many hidden layers and many units per layer are very flexible models with a very large number of parameters. As such, DNNs are challenging to optimize. To achieve real-time computation, embodiments disclosed herein enable fast DNN feature transformation via optimized memory bandwidth utilization. To optimize memory bandwidth utilization, a rate of accessing memory may be reduced based on a batch setting. A memory, corresponding to a selected given output neuron of a current layer of the DNN, may be updated with an incremental output value computed for the selected given output neuron as a function of input values of a selected few non-zero input neurons of a previous layer of the DNN in combination with weights between the selected few non-zero input neurons and the selected given output neuron, wherein a number of the selected few corresponds to the batch setting.

Fast deep neural network feature transformation via optimized memory bandwidth utilization

Growing community of inventors

Taintignies, Belgium

Christophe Ris