Granted Patent B2

US 12,373,033 · App. 18/153,337 · Granted Jul 29, 2025

Inventors: David Beattie (Bristol, GB); Rory Clark (Bristol, GB)

Assignee: ULTRAHAPTICS IP LTD

G06F3/016G06F3/011G06T7/42G06T7/44G06T7/45G06T2207/20056G06T2207/20081G06T2207/20084

Abstract

Claims (16)

1. A method of rendering roughness in visuo-haptic mid-air textures, comprising:

(1) forming a haptic prediction model by:

(a) training a visual texture dimension machine learning prediction and classification model;

(b) calculating a linear regression process to match visual texture dimension prediction values to validated specific haptic attributes to produce a particular haptic texture dimension;

(2) using the prediction and classification model for the visual texture dimension prediction values, developing an audio database and an associated rendering method to produce dynamic auditory feedback that is tied to visual features within a texture image; and

(3) converting image texture data to grayscale;

wherein training a visual texture dimension machine learning prediction and classification model comprises obtaining the image texture data based on the subjective observations of a plurality of test subjects;

wherein converting the image texture data to grayscale uses constant scaling;

calculating gray-level co-occurrence matrices for a plurality of displacement vectors; calculating matrices for correct distance values and pre-determined angles;

summing and averaging transpose matrices to produce a symmetric and semi-direction invariant matrix that is normalized so as to contain estimated probabilities for each pixel co-occurrence.

2. The method of claim 1 , wherein calculating a linear regression process uses a network with convolutional layers with Rectified Linear Unit activations and He normal kernel initializers.

3. The method of claim 2 , further comprising generating an output layer using a sigmoid activation function to output a predicted value of subjective roughness.

4. The method of claim 1 , wherein using the prediction and classification model for the visual texture dimension prediction values comprises converting a prediction value to a draw frequency of a haptic sensation using a linear regression approach.

5. The method of claim 1 , wherein dynamic auditory feedback that is tied to visual features within a texture image includes using audio intensity and frequency modulation based on variations in local features contained within the image texture.

6. The method of claim 5 , further comprising:

rendering the dynamic auditory feedback via parametric audio.

Assignments (4)

Continuity (4)

References Cited (400)

Cited By (1)

US 12,699,456