patent.fyi
Applications  /  US 2026/0354750
United States Patent Application Publication

US 2026/0354750 A1

Published
Nov 12, 2026
Filed
Oct 6, 2023
Earliest date on the face
Oct 7, 2022
Classification
A61B 5/4803
First published here
Nov 19, 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.

1.1Fixed analog scoring circuit

What ifWhat if a fixed analog circuit measures syllable timing and drives the speech assessment display?
Claim language“at least one processor”

A microphone feeds a band-pass filter, an envelope detector, and a comparator that emits a pulse for each syllable during a repeated-syllable task. A fixed timer gates the pulses for ten seconds, and a charge pump drives a meter calibrated in syllables per second.

1.2Optical sensing of throat motion

What ifWhat if an optical displacement sensor records throat surface motion while the subject sustains a vowel?
Claim language“a microphone”

A laser displacement sensor aimed at the front of the neck records surface motion at 2,000 samples per second. The processor measures the duration of periodic vibration and the variation in cycle length to report phonation stability.

1.3Silent articulation with facial electrodes

What ifWhat if the subject silently mouths a sequence of syllables during the assessment?
Claim language“of speaking aloud”

Surface electrodes near the lips and under the chin record muscle activity while the subject mouths a sequence shown on a display. The processor detects activation bursts and uses their timing and variability to score articulatory coordination.

1.3Waveform scoring without feature extraction

What ifWhat if the assessment model takes waveform samples directly and produces a score without a separate feature extraction stage?
Claim language“extract, from the audio data, digital biomarker data”

The microphone recording enters the model as a sequence of signed 16-bit samples collected at 16,000 samples per second. The model learns its internal representations during training and returns a speech function score directly from the sample sequence.

1.3Clinician interprets displayed measurements

What ifWhat if a clinician interprets the displayed measurements and enters the speech function rating?
Claim language“applying a speech function assessment model”

The device measures maximum sustained vowel duration, syllables per second, and the fraction of the recording occupied by pauses. The clinician reviews those values alongside audio playback and records the rating in a form.

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

Voice assessment for broadcast training

A training workstation prompts a presenter to describe a photograph aloud for 30 seconds and records the response through a headset microphone. Its processor estimates fundamental frequency across voiced frames and computes the median and interquartile range in semitones. A regression model trained on instructor ratings uses these features to report a pitch modulation score.

Component substitution

Piezoelectric throat microphone for vowel assessment

A piezoelectric contact microphone replaces the airborne microphone and mounts against the skin beside the larynx with an elastic strap. A high-impedance preamplifier conditions the signal before a converter samples it at 16 kHz. After prompting a sustained vowel, the processor estimates successive vocal periods and computes cycle-to-cycle period variation. A model calibrated on recordings from the contact microphone uses these measurements to report phonatory stability.

Parameter or scale variation

Repeated vowel trials with fixed timing

A display uses a color cue to prompt three sustained vowel trials of 5 seconds each with 10 seconds of rest between trials. A microphone records mono audio at 16 kHz, and the processor extracts a 2-second segment centered in each trial. The processor computes cepstral peak prominence for each segment and passes the median and range across trials to a model that reports voice periodicity.

Combination with existing systems

Speech assessment during remote rehabilitation

A speech assessment module runs alongside a video consultation system and receives a task selection from the clinician's session controls. The module presents a picture-description prompt and records directly from the patient device's microphone before the call audio encoder processes the signal. It estimates the first two vowel formants in voiced segments and passes their dispersion to a speech function model. The module sends the resulting score and recording quality indicators to the clinician's session dashboard.

Speech assessment with respiratory motion sensing

A respiratory inductance belt connects to the speech assessment device and measures changes in chest circumference while a microphone records prompted speech. The processor timestamps both streams against a common clock and aligns acoustic voicing onsets with changes in the belt signal. It computes the interval between the start of expiration and the start of voicing and the fraction of each expiratory phase occupied by voicing. A model uses these measurements to report respiratory and phonatory coordination.

Automated or data-driven variant

Automatic selection of speaking tasks

A tablet prompts the subject to sustain a vowel and records the response through its microphone. The processor extracts harmonicity and pitch variation and applies a trained model that estimates phonatory stability and the uncertainty of that estimate. When uncertainty exceeds a stored threshold, the tablet selects a follow-up task, such as a pitch glide or a repeated sustained vowel, based on which task best reduced uncertainty for similar recordings in a calibration dataset. The processor combines the features from both recordings to update the estimate.

Specific version of a part

Clinician delivers the speaking cue

The clinician starts recording with a foot switch and asks the subject to repeat a printed sentence. The processor detects the response interval, measures articulation rate and pause duration, and displays a speech function score.

Fixed point signal processor

A 32-bit signal processor reads digitized audio into a circular buffer. It computes short-time energy and autocorrelation in overlapping 25-millisecond frames.

Digital capacitive silicon microphone

A silicon diaphragm and a fixed electrode form a capacitor whose capacitance changes with sound pressure. The microphone package converts that change to a pulse-density bitstream, and the processor filters and decimates the stream to 16,000 samples per second.

Timed vowel task on screen

A display instructs the subject to sustain a vowel after a three-second visual countdown. The processor starts microphone capture at the end of the countdown and stops it after a ten-second recording window.

Ordinal regression for speech ratings

An ordinal regression model takes articulation rate, pause fraction, and pitch variability as inputs. Fitted coefficients and thresholds produce probabilities for four ordered speech function ratings.

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 2019/0172476 A1
Published
Jun 6, 2019
Filed
Dec 4, 2017
Class
G10L 21/0205
Document BUS 2022/0030113 A1
Published
Jan 27, 2022
Filed
Jul 21, 2021
Class
H04M 3/568
Claim 1
Document A
Document B
1.1at least one processor;
The electronic hardware may include micro-electronic-mechanical systems (MEMS), integrated circuits (e.g. application specific), microprocessors, microcontrollers, digital signal processors (DSPs), field programmable gate arrays (FPGAs), programmable logic devices (PLDs), gated logic, discrete hardware circuits, printed circuit boards (PCB) (e.g. flexible PCBs), and other suitable hardware configured to perform the various functionality described throughout this disclosure, e
B · US 2022/0030113 A1 · ¶ 106
1.2a microphone; and
Referring to FIG. 1, the multi-channel speech pickup may include a number of acoustic microphone signals that are produced by a number of acoustic microphones, respectively (not shown in FIG. 1 but see for example FIG. 4).
A · US 2019/0172476 A1 · ¶ 24
1.3a memory storing computer-readable instructions that, when executed by the at least one processor, cause the diagnostic device to: prompt the subject to perform a diagnostic task of speaking aloud; receive, via the microphone, audio data associated with the diagnostic task; extract, from the audio data, digital biomarker data associated with the speech function of the subject; and applying a speech function assessment model to the digital biomarker data, the speech function assessment model configured to generate an output indicative of the speech function of the subject.
In one aspect, computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method disclosed above.
B · US 2022/0030113 A1 · ¶ 14

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

Statements in these documents
There are thus a variety of acoustic environment conditions, which disturb an otherwise clean speech signal that is picked up by a microphone (as for example the voice of a near end talker in the room.) This hinders real-time applications such as voice trigger phrase detection, hands free telephony, and automatic speech recognition that may be performed upon the speech signal.
A · US 2019/0172476 A1 · ¶ 2
The process has low enough latency such that the enhanced speech signal can be used by a real-time speech processing application (such as voice trigger detection and ASR), with great accuracy (e.g., low false trigger rate and low word error rate.)
A · US 2019/0172476 A1 · ¶ 4
The description given herein may e.g. be useful in applications such as handsfree telephone systems, mobile telephones, teleconferencing systems (e.g. speakerphones), headphones, headsets, public address systems, gaming devices, karaoke systems, classroom amplification systems, etc.
B · US 2022/0030113 A1 · ¶ 2
Simply using a directional system with fixed directionality for all rooms is a suboptimal solution in most situations.
B · US 2022/0030113 A1 · ¶ 4
Cite aspatent.fyi, US 2026/0354750 A1, claim 1, first published 2026-11-19T03:31:31-05:00, https://patent.fyi/US20260354750/, SHA-256 43e9dbbc3796ffdcc7da7343a6f6eb53a49fe1020f0e42c9c596d20c1279cc36
First published
Permanent addresshttps://patent.fyi/US20260354750/
Content fingerprint (SHA-256)43e9dbbc3796ffdcc7da7343a6f6eb53a49fe1020f0e42c9c596d20c1279cc36
Methodv2026-11-18.2 · automated, not reviewed by a person
Correctionscorrections@patent.fyi