Voice denoising real-time interaction method based on cleaning robot

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a voice denoising real-time interaction method based on a cleaning robot, and the method comprises the steps: collecting a voice signal in a target environment through a preset cleaning robot; extracting a signal feature of the voice signal; performing noise suppression processing on noise in the voice signal according to the signal feature to obtain a primary de-noised signal of the voice signal; performing dynamic weighting on the primary de-noised signal by using a preset signal weighting algorithm and the signal features to obtain a weighted signal of the primary de-noised signal; generating an operation instruction of the cleaning robot according to the weighted signal; and generating a task sequence table of the cleaning robot according to the operation instruction, wherein the cleaning robot works according to the task sequence table. According to the invention, the efficiency of voice denoising real-time interaction based on the cleaning robot can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a real-time interactive method for voice denoising based on a cleaning robot. Background Art

[0002] The smart home market is developing rapidly, and more and more people are starting to use voice-controlled smart devices. The number of smart home devices around the world is increasing every year, and voice assistants and cleaning robots have become important components. Therefore, improving the voice recognition and real-time interaction efficiency of cleaning robots has become the key to improving user experience and market competitiveness.

[0003] The microphones and sensors that cleaning robots are usually equipped with have limited accuracy and sensitivity when picking up voice signals. Especially when the ambient noise is high, these sensors are easily interfered with, resulting in distortion and misjudgment of voice signals. In order to meet the growing demand for smart devices in modern homes and commercial environments, how to improve the efficiency of real-time interaction based on voice denoising for cleaning robots has become an urgent problem to be solved. Summary of the invention

[0004] The present invention provides a voice denoising real-time interaction method based on a cleaning robot, and its main purpose is to solve the problem of low efficiency in voice denoising real-time interaction based on a cleaning robot.

[0005] To achieve the above object, the present invention provides a real-time interactive method for voice denoising based on a cleaning robot, comprising: Using a preset cleaning robot to collect voice signals in the target environment; Extracting signal features of the speech signal; Performing noise suppression processing on the noise in the speech signal according to the signal characteristics to obtain a primary denoised signal of the speech signal; The primary denoised signal is dynamically weighted using a preset signal weighting algorithm and the signal feature to obtain a weighted signal of the primary denoised signal, wherein the preset signal weighting algorithm is: ; in, is the weighted signal of the primary denoised signal, is the power spectrum of the target signal in the primary denoised signal, is the power spectrum of the noise signal in the primary denoised signal, is the signal characteristic; generating an operation instruction of the cleaning robot according to the weighted signal; A task sequence table for the cleaning robot is generated according to the operation instruction, and the cleaning robot works according to the task sequence table.

[0006] Optionally, the collecting of voice signals in the target environment by using a preset cleaning robot includes: Determine a preset microphone array of the cleaning robot; Collecting an audio signal of a target environment according to the microphone array; The audio signal is level-adjusted to obtain a speech signal of the audio signal.

[0007] Optionally, the level-adjusting the audio signal to obtain a speech signal of the audio signal includes: Performing DC offset processing on the audio signal to obtain AC audio information of the audio signal; The communication audio information is normalized to obtain a speech signal of the audio signal.

[0008] Optionally, extracting the signal feature of the speech signal includes: Extracting frequency domain features of the speech signal; Extracting time domain features of the speech signal; The frequency domain features and the time domain features are aggregated into signal features of the denoised signal.

[0009] Optionally, extracting the frequency domain features of the speech signal includes: Performing short-time Fourier transform on the speech signal to obtain a transformed signal of the speech signal; The frequency domain features of the transformed signal are extracted, wherein the frequency domain features include: spectrum amplitude, spectrum phase, power spectrum density and spectrum impairment.

[0010] Optionally, extracting the time domain features of the speech signal includes: The short-time energy of the speech signal is generated according to the energy distribution of the speech signal, wherein the short-time energy is: ; in, is the short-time energy of the speech signal, is the speech signal, is the length of each time window, is the sampling index within the time window, Indicates the current moment; Generating a short-time zero crossing rate of the speech signal according to a change in the zero crossing point of the speech signal; The short-time energy and the short-time zero-crossing rate are collected as the time domain features of the speech signal.

[0011] Optionally, performing noise suppression processing on the noise in the speech signal according to the signal feature to obtain a primary denoised signal of the speech signal includes: The noise in the speech signal is subjected to noise suppression processing by using a preset denoising algorithm and the signal feature to obtain a primary denoising signal of the speech signal, wherein the preset denoising algorithm is: ; in, is the primary denoising signal of the speech signal, is the speech signal, is the estimated power spectrum of the noise in the speech signal.

[0012] Optionally, generating an operation instruction of the cleaning robot according to the weighted signal includes: Generate a semantic slot of the cleaning robot according to preset task parameters; Performing semantic analysis on the weighted signal, and mapping analysis results of the semantic analysis to the semantic slots one by one; The operation instructions of the cleaning robot are generated according to the mapping information of the semantic slots.

[0013] Optionally, generating a task sequence table for the cleaning robot according to the operation instruction includes: Parsing the operation instruction to obtain a parsed instruction of the operation instruction; Determining the instruction priority of the parsing instruction according to the instruction type of the parsing instruction; A task sequence table for the cleaning robot is generated according to the instruction priority and the parsed instruction.

[0014] Optionally, generating a task sequence table for the cleaning robot according to the instruction priority and the parsed instruction includes: The task sequence table of the cleaning robot is generated according to the preset sequence generation rule, the instruction priority and the parsing instruction, wherein the preset sequence generation rule is: ; in, is a sequence of operation steps in the task sequence table of the cleaning robot, It is A specific task instruction in the parsing instruction, is the instruction priority.

[0015] The present invention suppresses the noise of the collected voice signal and removes the environmental noise, so that the target voice signal is clearer, and the primary denoising signal is processed by dynamic weighting, and is further optimized according to the characteristics of the voice signal, and the weighting coefficient is intelligently adjusted according to the power spectrum of the target signal and the noise signal in the voice signal to further improve the voice quality. In the process of voice denoising, the combination of feature extraction and weighting enables the cleaning robot to dynamically adapt to different environmental conditions, generate accurate operation instructions according to clear voice input, and generate a corresponding task sequence table to guide the robot to efficiently perform cleaning tasks. Since the denoised voice instructions are more accurate, the robot is not easily disturbed by noise when performing tasks, and can complete the tasks as planned, thereby improving work efficiency. Therefore, the present invention proposes a real-time interactive method for voice denoising based on a cleaning robot, which can solve the problem of low efficiency of real-time interactive voice denoising based on a cleaning robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of a real-time interactive method for voice denoising based on a cleaning robot provided in one embodiment of the present invention; The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0017] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0018] The embodiment of the present application provides a real-time interactive method for voice denoising based on a cleaning robot. The execution subject of the real-time interactive method for voice denoising based on a cleaning robot includes but is not limited to at least one of the electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided in the embodiment of the present application. In other words, the real-time interactive method for voice denoising based on a cleaning robot can be executed by software or hardware installed on a terminal device or a server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services and artificial intelligence platforms.

[0019] Reference Figure 1 FIG. 1 is a flow chart of a real-time interactive method for voice denoising based on a cleaning robot provided by an embodiment of the present invention. In this embodiment, the real-time interactive method for voice denoising based on a cleaning robot includes: S1. Use a preset cleaning robot to collect voice signals in the target environment.

[0020] In an embodiment of the present invention, the method of collecting voice signals in a target environment by using a preset cleaning robot includes: Determine a preset microphone array of the cleaning robot; Collecting an audio signal of a target environment according to the microphone array; The audio signal is level-adjusted to obtain a speech signal of the audio signal.

[0021] In detail, according to the design of the cleaning robot, multiple microphones are built-in to form an array, and these microphones may be distributed in different positions of the robot, such as front, back, left, right, top, and bottom; Spatial positioning: ensure that the distance and angle between the microphones can provide sufficient spatial resolution to facilitate subsequent sound source localization and noise separation.

[0022] In detail, when the robot is started or before the first use, a self-calibration process is performed to ensure that all microphones are in normal working state and synchronize their timestamps, and the microphone gain settings are optimized to adapt to the current acoustic environment by collecting environmental background noise data. The audio signal of the target environment is collected according to the microphone array.

[0023] Furthermore, once the cleaning robot is started, the microphone array begins to capture sound signals from the surrounding environment in real time, including the user's voice commands and background noise, ensuring that the recorded audio quality is high enough so that subsequent processing steps can effectively distinguish between voice and other sounds.

[0024] In detail, the level adjustment of the audio signal to obtain the speech signal of the audio signal includes: Performing DC offset processing on the audio signal to obtain AC audio information of the audio signal; The communication audio information is normalized to obtain a speech signal of the audio signal.

[0025] Furthermore, the audio signal is normalized so that its average energy reaches a predetermined standard value, thereby ensuring that all speech signals have consistent volume characteristics.

[0026] Specifically, any possible DC component is eliminated to ensure that the audio signal contains only AC audio information, thus improving the accuracy of subsequent processing.

[0027] In detail, the audio signal is level-adjusted to maintain a consistent volume level regardless of the distance between the user and the robot.

[0028] In detail, the cleaning robot has multiple built-in microphones, which are arranged at different positions of the robot to collect sound signals in the surrounding environment; after the cleaning robot is started, it begins to collect sound signals in the environment in real time, including but not limited to user's voice commands, environmental noise, etc.

[0029] Furthermore, the collected voice signal is input into the signal processing system in the robot through a microphone for subsequent processing; the collected audio signal can be a single-channel or multi-channel signal (such as using multiple microphones for sound source localization and noise separation).

[0030] S2. Extracting signal features of the speech signal.

[0031] In an embodiment of the present invention, extracting the signal feature of the speech signal includes: Extracting frequency domain features of the speech signal; Extracting time domain features of the speech signal; The frequency domain features and the time domain features are aggregated into signal features of the denoised signal.

[0032] In detail, the extracting the frequency domain features of the speech signal includes: Performing short-time Fourier transform on the speech signal to obtain a transformed signal of the speech signal; The frequency domain features of the transformed signal are extracted, wherein the frequency domain features include: spectrum amplitude, spectrum phase, power spectrum density and spectrum impairment.

[0033] In detail, the spectrum amplitude represents the energy distribution of the signal at each frequency and is used to capture the main components of the signal; the spectrum phase represents the phase information of the signal at each frequency, and in the denoising process, the amplitude characteristics are usually focused on; the power spectrum density reflects the distribution of the signal power at different frequencies and is usually used to analyze the noise characteristics of the signal; the spectrum impairment refers to the evaluation index of the denoising effect obtained by comparing the difference between the denoised signal and the noise spectrum estimate.

[0034] In detail, short-time Fourier transform is to segment the signal and perform Fourier transform on each segment of the signal to obtain time-frequency analysis information, which can show the energy distribution of the signal at different times and frequencies.

[0035] In detail, the speech signal is subjected to STFT transformation to obtain a transformed signal, ie, a complex value in time and frequency, which represents the time-frequency characteristics of the signal.

[0036] In detail, the extracting the time domain features of the speech signal includes: The short-time energy of the speech signal is generated according to the energy distribution of the speech signal, wherein the short-time energy is: ; in, is the short-time energy of the speech signal, is the speech signal, is the length of each time window, is the sampling index within the time window, Indicates the current moment; Generating a short-time zero crossing rate of the speech signal according to a change in the zero crossing point of the speech signal; The short-time energy and the short-time zero-crossing rate are collected as the time domain features of the speech signal.

[0037] In detail, short-time energy reflects the energy distribution of the signal in each time window, helping to capture the intensity changes of the signal; short-time zero-crossing rate is used to measure the change of the zero-crossing point of the signal on the time axis, which is usually related to the noise characteristics of the signal.

[0038] In detail, Indicates the speech signal at time The short-term energy of a signal. It measures the energy of a signal within a time window; Represents the value of the speech signal at time The signal amplitude at Indicates the length of each time window, that is, at each moment , will take continuous sampling points to calculate the energy in the time window; is the sampling index within the time window, usually from 0 to , represents the different sampling points of the signal in the current time window; Indicates the current moment, used to calculate the point of the positioning signal.

[0039] In detail, short-time energy reflects the energy distribution of the signal in each time window. Short-time energy can help capture the intensity changes of the signal and usually helps to identify stronger signal components and weaker noise components during the denoising process.

[0040] In detail, the short-time zero-crossing rate is a measure of the characteristic of the change of the zero crossing point of the signal on the time axis, which indicates the volatility of the signal. The zero-crossing rate can be obtained by counting the number of times the signal passes through the zero point in each segment. The short-time zero-crossing rate is usually related to the noise signal. Noise usually causes more frequent zero crossings. Therefore, the change of the zero-crossing rate can help distinguish between signals and noise.

[0041] In detail, the frequency domain features and the time domain features are aggregated into the overall feature vector of the denoised signal. Specifically, the two types of features can be combined by simple concatenation or weighted averaging.

[0042] In general, signal feature extraction is an important step in speech denoising. By extracting frequency domain features (such as spectrum amplitude, spectrum phase, power spectrum density, spectrum impairment) and time domain features (such as short-time energy, short-time zero crossing rate), the characteristics of the signal can be fully analyzed to help perform denoising more effectively.

[0043] S3. Perform noise suppression processing on the noise in the speech signal according to the signal characteristics to obtain a primary denoised signal of the speech signal.

[0044] In an embodiment of the present invention, the step of performing noise suppression processing on the noise in the speech signal according to the signal feature to obtain a primary denoised signal of the speech signal includes: The noise in the speech signal is subjected to noise suppression processing by using a preset denoising algorithm and the signal feature to obtain a primary denoising signal of the speech signal, wherein the preset denoising algorithm is: ; in, is the primary denoising signal of the speech signal, is the speech signal, is the estimated power spectrum of the noise in the speech signal.

[0045] In the embodiment of the present invention, the noise suppression algorithm determines the best denoising strategy according to the frequency characteristics of the speech signal and the statistical characteristics of the noise. The signal after noise suppression is the result of preliminary denoising, that is, the primary denoising signal, and its noise interference has been greatly reduced.

[0046] In an embodiment of the present invention, The spectrum of the denoised speech signal, or the “primary denoised signal”; The spectrum of the original speech signal is usually represented in the frequency domain after a short-time Fourier transform (STFT) in time; Represents the power spectrum of the noise, which is an estimate of the noise spectrum; Represents the power spectrum of the original speech signal, that is, the energy of the signal at each frequency point.

[0047] In detail, the denoising method combines the noise power spectrum estimation and the spectrum information of the speech signal, and removes the noise by weighting the signal at each frequency point. Specifically: Power spectrum of the original signal Reflects the energy distribution of the speech signal at each frequency; the power spectrum of the noise It is an estimate of the noise, which helps to distinguish which frequency components are caused by noise and which are caused by speech.

[0048] Furthermore, for each frequency point, the ratio of the power spectrum of the signal to the power spectrum of the noise is calculated. This ratio can be understood as a factor that weights the importance of the signal. By weighting, the influence of the noise component is reduced. By adjusting the weight of the noise part, the impact of the noise is minimized and the main components of the speech signal are retained.

[0049] In detail, when the power spectrum of the signal is larger than the power spectrum of the noise (i.e., the speech signal is stronger and the noise is weaker), the denoising factor is close to 1 and the speech signal is basically unaffected; when the power spectrum of the noise is larger (i.e., the noise is stronger and the signal is weaker), the denoising factor approaches 0, the noise component is effectively suppressed, and the speech signal is retained.

[0050] In detail, the background noise can be effectively removed by weighting the adjusted spectrum information to obtain a primary denoised signal. Afterwards, the denoised spectrum signal can be restored to a time domain signal through operations such as inverse short-time Fourier transform to obtain the denoised speech.

[0051] In detail, the preset denoising algorithm can effectively reduce background noise and retain the clarity of the voice signal. It is especially suitable for situations where there is a large gap between signal and noise power. Through spectrum analysis, it can process the noise in each frequency band in a targeted manner. The noise in different frequency bands is suppressed to different degrees, thereby achieving a good denoising effect.

[0052] S4. Dynamically weight the primary denoised signal using a preset signal weighting algorithm and the signal feature to obtain a weighted signal of the primary denoised signal.

[0053] In the embodiment of the present invention, the cleaning robot weights the primary denoising signal according to the extracted signal features to enhance the effective components in the speech signal while suppressing the noise components.

[0054] Furthermore, the primary denoising signal is optimized by a signal weighting algorithm, which uses the power spectrum ratio of the target signal to the noise signal to dynamically adjust the signal weight. The core idea of ​​this step is to enable the robot to recognize voice commands more clearly and reduce noise interference through signal weighting.

[0055] In detail, the preset signal weighting algorithm is: ; in, is the weighted signal of the primary denoised signal, is the power spectrum of the target signal in the primary denoised signal, is the power spectrum of the noise signal in the primary denoised signal, is the signal characteristic.

[0056] In detail, the weighted signal represents a denoised signal obtained after weighting processing, and the spectrum of the signal is adjusted according to the power spectra of the target signal and the noise so as to highlight the target signal and suppress the noise component; is the target part of the signal (such as speech or other signals of interest), It is the power spectrum of the target signal at each frequency, which can help identify and increase the weight of the target signal; is the noise part of the signal, It is the power spectrum of the noise signal in the primary denoising signal, which is used to judge the strength of the noise component. The frequency region with stronger noise will get a lower weighting, thereby reducing the impact of noise on the signal; the signal feature represents the spectral information of the signal or other features related to denoising. It is used as the basic signal in the weighting process, combined with the previous spectrum analysis results.

[0057] Further, is a typical weighting factor used to adjust the weight according to the power spectrum ratio of the target signal and the noise signal. Power spectrum greater than the noise When , the weighting factor will approach 1, which means that the target signal will be less suppressed in this frequency region and the weight of the denoised signal will be higher.

[0058] In detail, when the power spectrum of the noise signal Greater than the power spectrum of the target signal When , the weighting factor will approach 0, which means that the noise component in this frequency area will be suppressed more, reducing the interference of noise on the signal.

[0059] Specifically, dynamic weighting is performed according to the power spectrum ratio of the target signal and noise at different frequencies, which can effectively distinguish the components of the signal and noise. The target signal will be enhanced in the frequency area with higher power, while the noise signal will be suppressed in the frequency area with higher power. The final output signal has a higher signal-to-noise ratio.

[0060] S5. Generate an operation instruction for the cleaning robot according to the weighted signal.

[0061] In an embodiment of the present invention, generating an operation instruction of the cleaning robot according to the weighted signal includes: Generate a semantic slot of the cleaning robot according to preset task parameters; Performing semantic analysis on the weighted signal, and mapping analysis results of the semantic analysis to the semantic slots one by one; The operation instructions of the cleaning robot are generated according to the mapping information of the semantic slots.

[0062] In detail, based on the weighted signals, the cleaning robot analyzes the command content through the voice recognition system and extracts the type of operation that needs to be performed (such as cleaning area, cleaning mode, path planning, etc.).

[0063] Furthermore, the process of generating the operation instructions of the cleaning robot based on the weighted signal is concretized into several key steps, including the generation and filling of semantic slots, and how to generate specific robot operation instructions based on these semantic slots.

[0064] In detail, semantic slots refer to some specific information or parameter locations defined according to the task requirements, environment, and actions that the robot can perform.

[0065] For example, when a cleaning robot performs a cleaning task, the semantic slot may contain the following information: cleaning area (such as "living room", "kitchen"); cleaning mode (such as "quick cleaning", "deep cleaning"); cleaning time (such as "10 minutes"); cleaning tools (such as "mopping" or "vacuuming").

[0066] Furthermore, according to the preset task parameters, the cleaning robot needs to identify and define these semantic slots. For example, the task parameters may be to clean the kitchen for 10 minutes, then the following semantic slots can be generated for the task: task slot: cleaning; area slot: kitchen; time slot: 10 minutes.

[0067] In detail, the input signal (such as voice commands or sensor data) is parsed to extract key task-related information, which is mapped into corresponding semantic slots.

[0068] For example, suppose the weighted signal is the voice command "Please clean the kitchen for 10 minutes". After semantic analysis, the system will recognize the following key information: Task: Cleaning; Area: Kitchen; Time: 10 minutes.

[0069] Specifically, the result of semantic analysis needs to be mapped to a specific slot according to the structure of the semantic slot. For example, if the result of semantic analysis is "clean the kitchen for 10 minutes", then the information will be mapped to the corresponding semantic slots in turn: task slot = cleaning; area slot = kitchen; time slot = 10 minutes. This mapping process is a way to structure information so that the robot can understand and process it.

[0070] Furthermore, based on this mapping information, the system can generate operating instructions for the cleaning robot. These operating instructions will tell the robot how to perform the task, such as: task type: cleaning; cleaning area: kitchen; cleaning time: 10 minutes. According to these instructions, the cleaning robot will start the corresponding cleaning mode and operation, such as: start cleaning mode, navigate to the kitchen area, set the cleaning time to 10 minutes, and start the cleaning task. This operation instruction generation process is a key step in the automation system, which enables the robot to make appropriate decisions and execution according to task requirements.

[0071] S6. Generate a task sequence table for the cleaning robot according to the operation instruction, and the cleaning robot works according to the task sequence table.

[0072] In an embodiment of the present invention, generating a task sequence table for the cleaning robot according to the operation instruction includes: Parsing the operation instruction to obtain a parsed instruction of the operation instruction; Determining the instruction priority of the parsing instruction according to the instruction type of the parsing instruction; A task sequence table for the cleaning robot is generated according to the instruction priority and the parsed instruction.

[0073] Specifically, the operating instructions of the cleaning robot need to be parsed, which is part of natural language processing, or parsing other forms of input instructions (such as graphical interface, touch screen instructions, etc.). The purpose of parsing is to extract key information from the instructions, such as: task type: such as cleaning, charging, maintenance, etc.; execution conditions: such as cleaning area, cleaning time, cleaning mode, etc.; priority: different tasks may have different priorities, and the system needs to judge the importance of the task. The parsed instructions may contain information such as "clean the living room", "fast mode" or "20 minutes cleaning time".

[0074] In detail, after parsing the instruction, the system needs to determine its priority according to the type of instruction. The priority of the instruction determines which operations the system should execute first when multiple tasks or subtasks are performed simultaneously.

[0075] For example: high-priority tasks, such as core cleaning tasks that must be completed, may be performed first; low-priority tasks, such as minor maintenance tasks or charging tasks, may be performed after completing other important tasks.

[0076] In detail, the priority setting can be based on the task type (such as cleaning tasks take precedence over maintenance tasks) or based on the user's customized settings (such as the user specifically requests that a certain area be cleaned first).

[0077] In detail, generating the task sequence table of the cleaning robot according to the instruction priority and the parsed instruction includes: The task sequence table of the cleaning robot is generated according to the preset sequence generation rule, the instruction priority and the parsing instruction, wherein the preset sequence generation rule is: ; in, is a sequence of operation steps in the task sequence table of the cleaning robot, It is A specific task instruction in the parsing instruction, is the instruction priority.

[0078] In detail, when generating the task sequence table, the system follows certain preset rules (sequence generation rules) that define the order in which tasks are to be executed. For example, tasks can be sorted based on their priority, time requirements, or resource availability.

[0079] Furthermore, the sorting methods may include: sorting by priority: tasks with higher priority are executed first; sorting by time requirements: tasks that need to be completed within the specified time are executed first; sorting by dependencies: some tasks may depend on the completion of other tasks (such as charging before cleaning), and these tasks will be sorted according to dependencies.

[0080] In detail, after receiving the instructions and their priorities, the system generates a task sequence table in order to determine the order and operation steps of task execution.

[0081] In the embodiment of the present invention, once the task sequence list is generated, the cleaning robot will perform the tasks in sequence according to the list.

[0082] In detail, the robot performs specific tasks (such as cleaning, charging, returning to the charging station, etc.); while performing tasks, the robot may feedback the current status to the system (such as task completion, current location, remaining time, etc.). If a task is not successfully completed for some reason (such as insufficient battery), the system can adjust subsequent tasks based on the feedback.

[0083] Overall, through this process, the system can ensure that the cleaning robots complete their tasks in the most effective and efficient order, and can cope with complex task scheduling requirements, such as prioritizing critical tasks when multiple tasks exist at the same time.

[0084] The present invention suppresses the noise of the collected voice signal and removes the environmental noise, so that the target voice signal is clearer, processes the primary denoising signal through dynamic weighting, further optimizes it according to the characteristics of the voice signal, and intelligently adjusts the weighting coefficient according to the power spectrum of the target signal and the noise signal in the voice signal to further improve the voice quality. In the process of voice denoising, the combination of feature extraction and weighting enables the cleaning robot to dynamically adapt to different environmental conditions, generate accurate operation instructions according to clear voice input, and generate a corresponding task sequence table to guide the robot to efficiently perform cleaning tasks. Since the denoised voice instructions are more accurate, the robot is not easily disturbed by noise when performing tasks, can complete tasks as planned, and improve work efficiency.

[0085] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0086] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method and technology of using digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A real-time interactive method for voice denoising based on a cleaning robot, characterized in that: The method comprises: Using a preset cleaning robot to collect voice signals in the target environment; Extracting signal features of the speech signal; Performing noise suppression processing on the noise in the speech signal according to the signal characteristics to obtain a primary denoised signal of the speech signal; The primary denoised signal is dynamically weighted using a preset signal weighting algorithm and the signal feature to obtain a weighted signal of the primary denoised signal, wherein the preset signal weighting algorithm is: ； in, is the weighted signal of the primary denoised signal, is the power spectrum of the target signal in the primary denoised signal, is the power spectrum of the noise signal in the primary denoised signal, is the signal characteristic; generating an operation instruction of the cleaning robot according to the weighted signal; A task sequence table for the cleaning robot is generated according to the operation instruction, and the cleaning robot works according to the task sequence table.

2. The real-time interactive method for voice denoising based on a cleaning robot according to claim 1, characterized in that: The method of collecting voice signals in the target environment by using a preset cleaning robot includes: Determine a preset microphone array of the cleaning robot; Collecting an audio signal of a target environment according to the microphone array; The audio signal is level-adjusted to obtain a speech signal of the audio signal.

3. The real-time interactive method for voice denoising based on a cleaning robot according to claim 2, characterized in that: The step of adjusting the level of the audio signal to obtain a speech signal of the audio signal includes: Performing DC offset processing on the audio signal to obtain AC audio information of the audio signal; The communication audio information is normalized to obtain a speech signal of the audio signal.

4. The real-time interactive method for voice denoising based on a cleaning robot according to claim 1, characterized in that: The extracting the signal feature of the speech signal comprises: Extracting frequency domain features of the speech signal; Extracting time domain features of the speech signal; The frequency domain features and the time domain features are aggregated into signal features of the denoised signal.

5. The real-time interactive method for voice denoising based on a cleaning robot according to claim 4, characterized in that: The extracting frequency domain features of the speech signal comprises: Performing short-time Fourier transform on the speech signal to obtain a transformed signal of the speech signal; The frequency domain features of the transformed signal are extracted, wherein the frequency domain features include: spectrum amplitude, spectrum phase, power spectrum density and spectrum impairment.

6. The real-time interactive method for voice denoising based on a cleaning robot according to claim 4, characterized in that: The extracting the time domain features of the speech signal comprises: The short-time energy of the speech signal is generated according to the energy distribution of the speech signal, wherein the short-time energy is: ； in, is the short-time energy of the speech signal, is the speech signal, is the length of each time window, is the sampling index within the time window, Indicates the current moment; Generating a short-time zero crossing rate of the speech signal according to a change in the zero crossing point of the speech signal; The short-time energy and the short-time zero-crossing rate are collected as the time domain features of the speech signal.

7. The real-time interactive method for voice denoising based on a cleaning robot according to claim 1, characterized in that: The performing noise suppression processing on the noise in the speech signal according to the signal characteristics to obtain a primary denoised signal of the speech signal includes: The noise in the speech signal is subjected to noise suppression processing by using a preset denoising algorithm and the signal feature to obtain a primary denoising signal of the speech signal, wherein the preset denoising algorithm is: ； in, is the primary denoising signal of the speech signal, is the speech signal, is the estimated power spectrum of the noise in the speech signal.

8. The real-time interactive method for voice denoising based on a cleaning robot according to claim 1, characterized in that: Generating the operation instruction of the cleaning robot according to the weighted signal comprises: Generate a semantic slot of the cleaning robot according to preset task parameters; Performing semantic analysis on the weighted signal, and mapping analysis results of the semantic analysis to the semantic slots one by one; The operation instructions of the cleaning robot are generated according to the mapping information of the semantic slots.

9. The real-time interactive method for voice denoising based on a cleaning robot according to claim 1, characterized in that: The step of generating a task sequence table for the cleaning robot according to the operation instruction comprises: Parsing the operation instruction to obtain a parsed instruction of the operation instruction; Determining the instruction priority of the parsing instruction according to the instruction type of the parsing instruction; A task sequence table for the cleaning robot is generated according to the instruction priority and the parsed instruction.

10. The real-time interactive method for voice denoising based on a cleaning robot according to claim 9, characterized in that: The step of generating a task sequence table for the cleaning robot according to the instruction priority and the parsed instruction includes: The task sequence table of the cleaning robot is generated according to the preset sequence generation rule, the instruction priority and the parsing instruction, wherein the preset sequence generation rule is: ； in, is a sequence of operation steps in the task sequence table of the cleaning robot, It is A specific task instruction in the parsing instruction, is the instruction priority.

Images