17946523. SELF-LEARNING NEUROMORPHIC ACOUSTIC MODEL FOR SPEECH RECOGNITION simplified abstract (Accenture Global Solutions Limited)

SELF-LEARNING NEUROMORPHIC ACOUSTIC MODEL FOR SPEECH RECOGNITION

Organization Name

Accenture Global Solutions Limited

Inventor(s)

Lavinia Andreea Danielescu of San Francisco CA (US)

Timothy M. Shea of Merced CA (US)

Kenneth Michael Stewart of Irvine CA (US)

Noah Gideon Pacik-nelson of Boston MA (US)

Eric Michael Gallo of Moretown VT (US)

SELF-LEARNING NEUROMORPHIC ACOUSTIC MODEL FOR SPEECH RECOGNITION - A simplified explanation of the abstract

This abstract first appeared for US patent application 17946523 titled 'SELF-LEARNING NEUROMORPHIC ACOUSTIC MODEL FOR SPEECH RECOGNITION

Simplified Explanation

The patent application describes a method for recognizing speech using a spiking neural network acoustic model implemented on a neuromorphic processor.

• The method involves receiving a set of feature coefficients representing acoustic energy from input audio, predicting speech sounds based on these coefficients, and updating the acoustic model using learning rules and predicted speech sounds.
• The acoustic model is implemented as a spiking neural network on a neuromorphic processor of a client device, with the processor updating parameters of the model to improve speech recognition accuracy.

Potential Applications

This technology could be applied in:

• Speech recognition systems
• Voice-controlled devices
• Language translation tools

Problems Solved

• Improved accuracy in speech recognition
• Efficient processing of audio data
• Real-time speech analysis

Benefits

• Faster and more accurate speech recognition
• Reduced computational resources required
• Enhanced user experience with voice-controlled devices

Potential Commercial Applications

Optimizing speech recognition technology for:

• Smart home devices
• Virtual assistants
• Customer service chatbots

Possible Prior Art

One possible prior art could be traditional speech recognition systems using deep learning models on conventional processors.

What are the limitations of using spiking neural networks for speech recognition?

Spiking neural networks may require more complex training algorithms and computational resources compared to traditional neural networks, which could limit their practical implementation in some applications.

How does the neuromorphic processor improve the efficiency of speech recognition compared to conventional processors?

The neuromorphic processor is designed to mimic the structure and function of the human brain, allowing for parallel processing and low-power consumption, which can enhance the speed and accuracy of speech recognition tasks.

Original Abstract Submitted

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for recognizing speech using a spiking neural network acoustic model implemented on a neuromorphic processor are described. In one aspect, a method includes receiving, a trained acoustic model implemented as a spiking neural network (SNN) on a neuromorphic processor of a client device, a set of feature coefficients that represent acoustic energy of input audio received from a microphone communicably coupled to the client device. The acoustic model is trained to predict speech sounds based on input feature coefficients. The acoustic model generates output data indicating predicted speech sounds corresponding to the set of feature coefficients that represent the input audio received from the microphone. The neuromorphic processor updates one or more parameters of the acoustic model using one or more learning rules and the predicted speech sounds of the output data.