patent.fyi
Applications  /  US 2026/0354352
United States Patent Application Publication

US 2026/0354352 A1

Published
Nov 12, 2026
Filed
Nov 29, 2023
Earliest date on the face
Nov 30, 2022
Classification
A01K 29/005
First published here
Nov 18, 2026

Generated automatically. Not reviewed by a person. Not legal advice or an opinion.

Section 1 · machine-generated

What if?

Questions about how individual parts of claim 1 might be changed. A question may describe something different from what the claim language recites, or a version of it; no view is offered on which.

1.1One microphone per enclosure

What ifWhat if one microphone monitors the enclosure?
Claim language“a plurality of sensors”

A ceiling-mounted microphone collects sound from a single calf pen. A processor analyzes the single audio stream for repeated coughs and reports events at the pen level.

1.2Caretaker enters observed events

What ifWhat if a caretaker enters observations through a keypad?
Claim language“receiving, from the sensor, a signal”

A caretaker selects a pen and enters the number of coughs heard during a five-minute observation. The terminal records the pen identifier, count, and observation time for review.

1.3Direct waveform input

What ifWhat if the processor feeds received waveform samples directly into the neural network?
Claim language“a transform on the received signal”

A microphone digitizer supplies consecutive blocks of 16,000 signed samples at 16 kHz. The processor passes each block unchanged to the network input, where one-dimensional convolutional layers detect short patterns in the waveform.

1.4Cough detection by thresholds

What ifWhat if fixed rules detect repeated coughing?
Claim language“activity of interest neural network processing”

The processor computes short-time spectral energy and identifies bursts that exceed a background threshold across multiple frequency bands. It checks burst duration and raises a pen alert when the count reaches a configured limit within five minutes.

1.4Cough detection by template matching

What ifWhat if the processor compares audio events with stored cough recordings?
Claim language“activity of interest neural network processing”

The processor extracts a sequence of spectral features from each sound burst and compares it with feature sequences from labeled cough recordings. Dynamic time warping aligns sequences of different durations, and a distance threshold determines which bursts count as coughs.

Section 2 · machine-generated

Variations

Further arrangements of the same subject matter, described here as of the publication date at the foot of this page.

Other field of use

Abnormal swimming detection in hatchery tanks

Wall-mounted hydrophones in fish hatchery tanks measure underwater sound associated with swimming and surface activity. A shared processor converts each waveform into a spectrogram and runs a neural network trained on recordings labeled for sustained surface thrashing. The processor sends the tank identifier, event time, and confidence score to a server when the score exceeds a configured threshold.

Component substitution

Wavelet features for barn sound classification

Microphones in livestock pens send digitized waveforms to a shared processor. The processor uses a discrete wavelet packet transform in place of a Fourier transform and computes the energy in each terminal band for successive time windows. A neural network classifies the resulting band-energy sequences for events such as repeated coughing, and the processor sends each detected event and its microphone identifier to a server.

Floor vibration sensing for gait changes

Sealed accelerometers mounted under a livestock alley floor replace airborne microphones and measure vibration from footfalls. A shared processor divides each accelerometer waveform into animal passage segments using entrance and exit beam sensors, then transforms the segments into time-frequency maps. A neural network scores the maps for irregular footfall timing associated with altered gait, and the processor reports flagged passages to a server with the alley identifier and passage time.

Parameter or scale variation

Overlapping audio windows across twelve pens

Twelve microphones serve twelve pens and deliver separate channels to a shared processor at 16,000 samples per second per channel. The processor computes a log-mel spectrogram for each four-second window and advances the window by one second at a time. It runs a neural network on each spectrogram and reports a welfare event with the pen identifier and onset time after three consecutive windows exceed a confidence threshold of 0.8 for the same event class.

Combination with existing systems

Welfare alerts linked to milking records

A microphone above each stall in an automated milking system sends audio to a shared processor, and the stall reader supplies the animal identifier and occupancy times. The processor transforms the audio into spectrograms and uses a neural network to detect coughing during each occupied interval. It sends the event time, animal identifier, and confidence score to the herd management server, which links the event to the corresponding milking record.

Automated or data-driven variant

Alert thresholds based on equipment state

Microphones in a poultry house send audio to a processor that also reads the current ventilation fan stage from the ventilation controller. The processor computes a spectrogram for each audio window and runs a neural network to score distress calling. Software derives a separate alert threshold for each fan stage from the 99th percentile of scores in recordings labeled as routine activity. The processor applies the threshold for the current stage and sends the event class, time, and sensor identifier to a server when the score exceeds it.

Specific version of a part

One microphone in each pen

A directional microphone mounted above each pen points toward the animal area. Each channel carries a pen identifier so the processor associates detected events with the corresponding enclosure.

Logarithmic filter bank features

The processor computes a short-time Fourier transform on 25-millisecond frames with a 10-millisecond hop. It sums spectral power into 64 frequency bands spaced on a perceptual scale and takes the logarithm of each band value.

Convolutional classifier for coughs

A compact convolutional neural network processes one-second spectrogram segments and outputs a cough probability for each segment. The processor counts detections above a configured probability threshold and flags the pen when its five-minute count reaches a configured limit.

Buffered event reports over ethernet

The monitoring unit sends the pen identifier, event type, timestamp, and event count to a server over wired Ethernet. It stores each report in local flash memory until the server acknowledges receipt.

Section 3 · quoted from USPTO records

Other work

Passages reproduced word for word from two earlier-published U.S. patent documents, placed beside the language of claim 1. No characterization is offered. Each passage is marked A or B for the document it comes from; ¶ numbers are the paragraph numbers in the USPTO’s published full text of that document.

Document AUS 8,831,936 B2
Published
Sep 9, 2014
Filed
May 28, 2009
Class
G10L 21/02
Document BUS 2022/0051687 A1
Published
Feb 17, 2022
Filed
Dec 18, 2019
Class
G10L 25/24
Claim 1
Document A
Document B
1.1for each of a plurality of sensors and
In a typical application of apparatus A100, each channel of sensed audio signal S10 is based on a signal from a corresponding one of an array of M microphones, where M is an integer having a value greater than one.
A · US 8,831,936 B2 · ¶ 163
1.2receiving, from the sensor, a signal indicative of input sensed by the sensor relating to an animal;
In a typical application of apparatus A100, each channel of sensed audio signal S10 is based on a signal from a corresponding one of an array of M microphones, where M is an integer having a value greater than one.
A · US 8,831,936 B2 · ¶ 163
1.3performing, by a processor, a transform on the received signal;
Apparatus A132 also includes a transform module TR10 that is configured to transform source signal S20 and noise reference S30 into the transform domain. In a typical example, transform module TR10 is configured to perform a fast Fourier transform (FFT), such as a 128-point, 256-point, or 512-point FFT, on each of source signal S20 and noise reference S30 to produce the respective frequency-domain signals.
A · US 8,831,936 B2 · ¶ 174
A sound processing apparatus comprising: a memory storing processing instructions; and at least one processor configured to execute the processing instructions, the processing instructions comprising performing a Fourier transform and then a cepstral analysis of a sound signal and extracting, as feature values of the sound signal, values including frequency components obtained by the Fourier transform of the sound signal and a value based on a result obtained by the cepstral
B · US 2022/0051687 A1 · claim 8
1.4performing, by the processor, activity of interest neural network processing on the transformed signal to determine if an activity of interest has occurred, the activity of interest occurring if the sensed input is indicative of ill health and/or poor welfare of the animal; and
1.5transmitting an indication of the activity of interest to a server.

An empty cell means only that no passage was reproduced there.

Statements in these documents
For applications in which communication occurs in noisy environments, it may be desirable to separate a desired speech signal from background noise.
A · US 8,831,936 B2 · ¶ 7
A noisy acoustic environment may also tend to mask, or otherwise make it difficult to hear, a desired reproduced audio signal, such as the far-end signal in a phone conversation.
A · US 8,831,936 B2 · ¶ 8
However, an impulse sound, such as a sound that occurs when a ceiling light or a home appliance is switched on or a sound that occurs when a door is closed, shows approximately flat frequency characteristics in a certain range and therefore features thereof are difficult to grasp.
B · US 2022/0051687 A1 · ¶ 8
Accordingly, an object of the present invention is to provide a sound processing method, sound processing apparatus, and program that are able to resolve the difficulty in recognizing an impulse sound.
B · US 2022/0051687 A1 · ¶ 9
Cite aspatent.fyi, US 2026/0354352 A1, claim 1, first published 2026-11-18T21:55:14-05:00, https://patent.fyi/US20260354352/, SHA-256 f2173403d22f1ea558c04fc6e5b1174f6b64db8022d6d6309f7d82b1f6e82a60
First published
Permanent addresshttps://patent.fyi/US20260354352/
Content fingerprint (SHA-256)f2173403d22f1ea558c04fc6e5b1174f6b64db8022d6d6309f7d82b1f6e82a60
Methodv2026-11-18.2 · automated, not reviewed by a person
Correctionscorrections@patent.fyi