Speech control of a medical apparatus

Patent metadata

Field	Value
Publication number	US-12469492-B2
Application number	US-202217949540-A
Country	US
Kind code	B2
Filing date	Sep 21, 2022
Priority date	Sep 23, 2021
Publication date	Nov 11, 2025
Grant date	Nov 11, 2025

Abstract

One or more example embodiments provides a method for speech control of a medical apparatus, a corresponding speech control apparatus, a medical system comprising the speech control apparatus, a computer program product and a computer-readable storage medium.

First claim

The invention claimed is:

1 . A method for speech control of a medical apparatus, the method comprising:
capturing an audio signal containing operator speech input for controlling the medical apparatus; first analyzing the audio signal to obtain a first speech analysis result; recognizing a first speech command from the operator based on the first speech analysis result; ascertaining a verification signal for the first speech command, the ascertaining including,
second analyzing the audio signal to obtain a second speech analysis result, the second analyzing including applying a computational linguistics algorithm to the audio signal, and the computational linguistics algorithm being configured to only recognize safety-critical speech commands,
recognizing a second speech command from the operator based on the second speech analysis result, and
comparing the first speech command and the second speech command to obtain a comparison result,
the verification signal confirming the first speech command based on the comparison result satisfying a conformity criterion; generating a control signal for controlling the medical apparatus based on the verification signal, the control signal being configured to control the medical apparatus according to the first speech command; and inputting the control signal into the medical apparatus.

2 . The method of claim 1 , wherein
the computational linguistics algorithm is a first computational linguistics algorithm; the first analyzing includes applying a second computational linguistics algorithm to the audio signal, the second computational linguistics algorithm including a first trained function; the first computational linguistics algorithm including a second trained function; and the first trained function and the second trained function are different from one another.

3 . The method of claim 2 , wherein the second trained function is configured only to recognize the safety-critical speech commands.

4 . The method of claim 3 , wherein
the first analyzing includes tokenizing for segmenting at least one segmenting letters, words or sentences within the audio signal via the second computational linguistics algorithm to obtain first tokenization information, the first speech analysis result including the first tokenization information; and the second analyzing includes tokenizing for segmenting at least one of letters, words or sentences within the audio signal via the first computational linguistics algorithm to obtain second tokenization information, the second speech analysis result including the second tokenization information.

5 . The method of claim 4 , wherein
the first analyzing includes semantic analysis of the audio signal via the second computational linguistics algorithm to obtain first semantic information, the first speech analysis result including the first semantic information; and the second analyzing includes semantic analysis of the audio signal via the first computational linguistics algorithm to obtain second semantic information, the second speech analysis result including the second semantic information.

6 . The method of claim 5 , wherein the first trained function is configured to use at least one of the second tokenization information or the second semantic information to identify at least one of:
structurally unique speech commands as the safety-critical speech commands, or lexically unique speech commands as the safety-critical speech commands.

7 . The method of claim 5 , wherein the first trained function is configured to identify a speech command with at least three syllables as one of the safety-critical speech commands.

8 . The method of claim 2 , further comprising:
classifying the first speech command into one among of a plurality of safety classes, at least one first safety class among the plurality of safety classes corresponding to the safety-critical speech commands, wherein the ascertaining of the verification signal for the first speech command is performed based on the first speech command being classified in the at least one first safety class.

9 . The method of claim 1 , wherein the capturing of the audio signal includes:
capturing the audio signal via a first input apparatus and a second input apparatus, the first input apparatus and the second input apparatus being different from one another, wherein
the first analyzing is performed using the audio signal captured via the first input apparatus, and
the second analyzing is performed using the audio signal captured via the second input apparatus.

10 . A non-transitory computer-readable storage medium storing program sections that, when executed by a system, cause the system to perform the method of claim 1 .

11 . The method of claim 1 , wherein
the computational linguistics algorithm is a first computational linguistics algorithm; the first analyzing includes tokenizing for segmenting at least one of letters, words or sentences within the audio signal via a second computational linguistics algorithm to obtain first tokenization information, the first speech analysis result including the first tokenization information; and the second analyzing includes tokenizing for segmenting at least one of letters, words or sentences within the audio signal via the first computational linguistics algorithm to obtain second tokenization information, the second speech analysis result including the second tokenization information.

12 . The method of claim 2 , wherein
the first analyzing includes tokenizing for segmenting at least one of letters, words or sentences within the audio signal via the second computational linguistics algorithm to obtain first tokenization information, the first speech analysis result including the first tokenization information; and the second analyzing includes tokenizing for segmenting at least one of letters, words or sentences within the audio signal via the first computational linguistics algorithm to obtain second tokenization information, the second speech analysis result including the second tokenization information.

13 . The method of claim 3 , further comprising:
classifying the first speech command into one among a plurality of safety classes, at least one first safety class among the plurality of safety classes corresponding to the safety-critical speech commands, wherein the ascertaining of the verification signal for the first speech command is performed based on the first speech command being classified in the at least one first safety class.

14 . The method of claim 4 , further comprising:
classifying the first speech command into one among a plurality of safety classes, at least one first safety class among the plurality of safety classes corresponding to the safety-critical speech commands, wherein the ascertaining of the verification signal for the first speech command is performed based on the first speech command being classified in the at least one first safety class.

15 . The method of claim 5 , further comprising:
classifying the first speech command into one among a plurality of safety classes, at least one first safety class among the plurality of safety classes corresponding to the safety-critical speech commands, wherein the ascertaining of the verification signal for the first speech command is performed based on the first speech command being classified in the at least one first safety class.

16 . The method of claim 2 , wherein
the first analyzing includes semantic analysis of the audio signal via the second computational linguistics algorit

Assignees

• Siemens Healthineers Ag

Inventors

• Kuhrt Soeren
• Rothbauer Stefan
• Kilian Lennart

Classifications

• G10L2015/223

  Execution procedure of a spoken command · CPC title

• A61B2560/0493

  controlled by voice · CPC title

• G10L15/22 Primary

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title