Sound localization with artificial neural network
The location of a sound within a given spatial volume may be used in applications such as augmented reality environments. An artificial neural network processes time-difference-of-arrival data (TDOA) from a known microphone array to determine a spatial location of the sound. The neural network may be located locally or available as a cloud service. The artificial neural network is trained with perturbed and non-perturbed TDOA data.
1. A system comprising:
a plurality of microphones configured in a pre-determined arrangement;
a time-difference-of-arrival module configured to determine relative time-difference-of-arrival of an acoustic signal at the plurality of microphones; and
a trained artificial neural network module configured to accept the determined time-difference-of-arrival and generate spatial coordinates of the acoustic source.
2. The system of claim 1 , wherein the neural network comprises a hidden Markov model.
3. The system of claim 1 , further comprising supervised training of the neural network with backpropagation using acoustic signals at known spatial coordinates.
4. The system of claim 1 , further comprising a network interface module configured to access a remote cloud resource comprising a second trained neural network configured to provide higher accuracy spatial coordinates compared to the trained artificial neural network.
5. The system of claim 1 , wherein the training of the trained neural network comprises:
acquiring time-difference-of-arrival data associated with an acoustic source at a known spatial location;
perturbing the time-difference-of-arrival data within a pre-determined range; and
associating within the neural network the perturbed and non-perturbed data with the known spatial location.
6. The system of claim 1 , wherein the neural network is implemented within a dedicated chip.
7. The system of claim 1 , wherein the acoustic signal comprises human speech.
8. The system of claim 1 , wherein a user at least in part generates the acoustic signal.
9. The system of claim 1 , wherein the acoustic signal comprises an audible gesture within an augmented reality environment.
10. One or more computer-readable media storing computer-executable instructions that, when executed, cause one or more processors to perform acts comprising:
acquiring time-difference-of-arrival data associated with an acoustic source at known spatial coordinates relative to a pre-determined microphone configuration of a plurality of microphones;
perturbing the time-difference-of-arrival data within a pre-determined range;
receiving the perturbed time-difference-of-arrival data at a plurality of artificial neural network input nodes; and
training the artificial neural network to associate the perturbed time-difference-of-arrival data with the known spatial coordinates.
11. The one or more computer-readable storage media of claim 10 , further comprising:
receiving non-perturbed time-difference-of-arrival data at the plurality of artificial neural network input nodes; and
training the artificial neural network to associate the non-perturbed time-difference-of-arrival data with the known spatial coordinates.
12. The one or more computer-readable storage media of claim 10 , wherein artificial neural network comprises a hidden Markov model.
13. The one or more computer-readable storage media of claim 10 , the training further comprising supervised learning with backpropagation.
14. The one or more computer-readable storage media of claim 10 , the training further comprising unsupervised learning.
15. One or more computer-readable media storing computer-executable instructions that, when executed, cause one or more processors to perform acts comprising:
acquiring time-difference-of-arrival data associated with an acoustic source;
receiving the time-difference-of-arrival data at a plurality of artificial neural network nodes in a trained artificial neural network; and
processing the time-difference-of-arrival data with the trained artificial neural network configured to generate at a plurality of output nodes spatial coordinates of the acoustic source.
16. The one or more computer-readable storage media of claim 15 , wherein the processing further comprises a plurality of artificial neural networks configured such that:
a first trained artificial neural network processes the time-difference-of-arrival data locally; and
a second trained artificial neural network at a remote location, the second trained artificial neural network comprising additional nodes compared to the first trained artificial neural network.
17. The one or more computer-readable storage media of claim 16 , wherein the second trained artificial neural network is configured to execute as a cloud compute resource accessible to a plurality of users.
18. The one or more computer-readable storage media of claim 16 , further comprising modifying the spatial coordinates of the acoustic source with combined results from the first and second trained neural networks.
19. One or more computer-readable media storing computer-executable instructions that, when executed, cause one or more processors to perform acts comprising:
acquiring perturbed and non-perturbed time-difference-of-arrival data associated with an acoustic source in a gaseous environment;
receiving the time-difference-of-arrival data at a plurality of artificial neural network input nodes of a trained artificial neural network; and
accepting the time-difference-of-arrival at the plurality of artificial neural network input nodes and generating spatial coordinates of the acoustic source.
20. The one or more computer-readable storage media of claim 19 , further comprising updating the trained artificial neural network with spatial coordinates of the acoustic source determined by a ranging system.
21. The one or more computer-readable storage media of claim 20 , wherein the ranging system comprises a structured light module.