Smart home equipment, control method thereof and smart home system

Abstract

The invention relates to smart home equipment, a control method thereof and a smart home system, and relates to the technical field of smart home, and the method comprises the following steps: capturing a user voice instruction, obtaining an original audio signal, and carrying out noise reduction and filtering to obtain a clear voice segment; after the control text content is converted into control text content, keyword segmentation is carried out, and equipment identification words and operation instruction words are extracted; searching a target equipment communication address in a local equipment mapping table according to the equipment identification word, and generating a structured control command frame in combination with the operation instruction word; coding and modulating the frame through a wireless communication module, generating a radio frequency control signal and sending the radio frequency control signal to a home gateway; and the gateway forwards the signal according to the communication address of the target equipment, and controls the target intelligent equipment to execute the preset function action matched with the operation instruction word, so that the technical problems that the system is often difficult to accurately segment keywords and correctly match the target equipment and the operation action, and the error control or missing control phenomenon is easy to occur are solved.

Description

Technical Field

[0001] This invention relates to the field of smart home technology, and in particular to a smart home device and its control method, and a smart home system. Background Technology

[0002] With the rapid development of IoT technology, smart home systems are gradually entering thousands of households, bringing unprecedented convenience and comfort to people's lives. Among the many smart interaction methods, voice control, due to its natural and intuitive operation, has become the main bridge for communication between users and smart home devices. Traditional smart home control mostly relies on mobile applications or physical remote controls, which are cumbersome and not user-friendly. The introduction of voice control technology has greatly improved the user experience, making "hands-free" a reality. Currently, mainstream voice control systems typically rely on cloud servers for voice recognition and semantic understanding. Although the recognition accuracy is high, it is highly dependent on the network environment, resulting in problems such as high response latency and service stability affected by network fluctuations.

[0003] More importantly, existing voice control methods still fall short in terms of the real-time performance and accuracy of command parsing and execution. When users issue voice commands involving multiple devices or complex operations, the system often struggles to accurately segment keywords and correctly match the target device with the operation, leading to mis-control or missed control. Furthermore, control commands often use common wireless protocols during transmission, lacking optimization mechanisms for efficient communication between the home gateway and terminal devices. This results in low signal modulation and transmission efficiency, affecting the overall system's response speed and stability. Therefore, there is an urgent need for a smart home voice control method that can perform local voice processing, possesses high noise immunity, and supports accurate multi-device recognition and rapid control to address the core issues of high latency, significant privacy risks, poor compatibility, and inaccurate response in current technologies. Summary of the Invention

[0004] The purpose of this invention is to at least partially solve one of the technical problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides a control method for smart home devices, comprising the following steps:

[0006] The user's voice commands are captured to obtain the original audio signal, and the original audio signal is then subjected to noise reduction filtering to obtain a clear voice segment.

[0007] The clear audio segment is converted into control text content, and the control text content is segmented by keywords to obtain device identification words and operation instruction words;

[0008] Based on the device identifier, the corresponding target device communication address is found in the local device mapping table, and a structured control command frame is generated based on the operation instruction word;

[0009] The structured control command frame is encoded and modulated to obtain a radio frequency control signal, which is then sent to the home gateway. The home gateway, based on the target device's communication address, sends and controls the target smart device to perform a preset function action that matches the operation command word.

[0010] Furthermore, the user's voice commands are captured to obtain the original audio signal, and the original audio signal is then subjected to noise reduction filtering to obtain clear speech segments, including:

[0011] A sound wave sensor is used to capture the sound waves of the user's voice commands to obtain the raw audio signal, and the raw audio signal is subjected to spectrum analysis to obtain a spectrum distribution map;

[0012] Based on the spectrum distribution map, the original audio signal is subjected to noise frequency band identification to obtain the noise frequency band range. Then, the signal components in the original audio signal that are within the noise frequency band range are filtered out by the frequency band filter to obtain the preliminary noise-reduced audio signal.

[0013] The residual noise of the initial noise-reduced frequency signal is evaluated to obtain the residual noise intensity, and the gain of the initial noise-reduced frequency signal is adaptively adjusted based on the residual noise intensity to obtain a clear speech segment.

[0014] Furthermore, based on the aforementioned spectral distribution map, noise frequency bands are identified in the original audio signal to obtain the noise frequency band range, including:

[0015] Energy peak detection is performed on the spectrum distribution map to obtain a set of energy peaks, and the energy peak set is sorted by amplitude to obtain a sorted peak list; based on the sorted peak list, the original audio signal is divided into frequency bands to obtain multiple candidate frequency bands, and the energy proportion of each candidate frequency band is calculated to obtain the energy proportion value of each frequency band; the energy proportion values ​​of each frequency band are filtered by a preset noise threshold to obtain the noise frequency band range.

[0016] Furthermore, converting the clear speech segment into control text content includes:

[0017] Speech features are extracted from the clear speech segment to obtain a speech feature sequence, and acoustic unit matching is performed on the speech feature sequence to obtain an acoustic unit sequence.

[0018] Based on the acoustic unit sequence, a pronunciation mapping is performed to obtain a pinyin symbol string, and the pinyin symbol string is segmented into continuous syllables to obtain syllable unit groups;

[0019] The syllable unit group is matched with words to obtain a candidate word set, and the candidate word set is verified for contextual association to obtain the control text content.

[0020] Furthermore, the control text content is segmented by keywords to obtain device identification words and operation instruction words, including:

[0021] The control text content is initially segmented into character blocks to obtain a set of character blocks. Part-of-speech tagging is then performed on the character block set to obtain a part-of-speech tag set. Based on the part-of-speech tag set, the character block set is further divided into semantic units to obtain semantic unit groups. Device lexicon matching is performed on the semantic unit groups to obtain a set of candidate device words. Contextual association verification is then performed on the candidate device word set to obtain device identifier words. After removing word blocks containing device identifier words, operation instruction lexicon matching is performed on the remaining semantic unit groups in the semantic unit groups to obtain a set of candidate operation words. Functional logic verification is then performed based on the device identifier words and the set of candidate operation words to obtain operation instruction words.

[0022] Furthermore, based on the device identifier, the corresponding target device communication address is searched from the local device mapping table, including:

[0023] The local device mapping table is parsed to obtain a device information list, and the device identifier field and communication address field in the device information list are associated and marked to obtain an association mapping table;

[0024] The device identifier is processed by character standardization to obtain a standard identifier, and a field matching query is performed on the association mapping table based on the standard identifier to obtain candidate communication addresses;

[0025] The candidate communication addresses are counted to obtain the address count result, and the uniqueness of the candidate communication addresses is verified based on the address count result to obtain the target device communication address.

[0026] Furthermore, the structured control command frame is signal encoded and modulated to obtain a radio frequency control signal, including:

[0027] The structured control command frame is parsed to obtain a digital instruction sequence, and the frame header and frame tail of the digital instruction sequence are checked to obtain the checked digital instruction.

[0028] Based on the wireless communication protocol, the verified digital instructions are subjected to symbol mapping conversion to obtain a set of modulation symbols, and the set of modulation symbols is optimized by constellation diagram distribution to obtain optimized modulation symbols;

[0029] The optimized modulation symbol is modulated by carrier frequency using radio frequency modulation technology to obtain a preliminary radio frequency control signal. The power spectral density of the radio frequency control signal is then detected to obtain a radio frequency control signal that conforms to the transmission standard.

[0030] Furthermore, constellation diagram distribution optimization is performed on the modulation symbol set to obtain optimized modulation symbols, including:

[0031] The initial constellation points are located in the set of modulation symbols to obtain an initial constellation point position set, and the distance between neighboring points is calculated in the initial constellation point position set to obtain a neighboring point distance set.

[0032] Based on the set of distances to neighboring points, the initial set of constellation point positions is filtered for outliers to obtain anomaly constellation point summation. The positions of the outlier constellation points are then corrected by moving them toward the center of their neighboring normal constellation points to obtain the corrected constellation point positions.

[0033] The modulation symbol set is remapped based on the corrected constellation point positions to obtain optimized modulation symbols.

[0034] Furthermore, the radio frequency control signal is sent to the home gateway, and the home gateway, based on the target device's communication address, sends and controls the target smart device to execute preset functional actions matching the operation command words, including:

[0035] The radio frequency control signal is encapsulated with a protocol header to obtain a radio frequency signal with a protocol header; an address field is embedded in the radio frequency signal with the protocol header based on the communication address of the target device to obtain an address-bearing radio frequency signal; the address-bearing radio frequency signal is sent to the home gateway, and a routing path is planned for the address-bearing radio frequency signal based on the routing table of the home gateway to obtain planned path information; the address-bearing radio frequency signal is sent to the target smart device through the wireless transmission module of the home gateway according to the planned path information;

[0036] The address-bearing radio frequency signal is frame-parsed based on the protocol parsing engine technology in the target smart device to extract the operation instruction words in the address-bearing radio frequency signal, and the operation instruction words are encoded and converted based on the function mapping table stored in the home gateway to obtain the preset function code.

[0037] The preset function code is sent to the target smart device corresponding to the device matching result, and the target smart device is controlled to perform the action corresponding to the preset function code.

[0038] The present invention also provides a smart home system, comprising:

[0039] The capture module is used to capture sound waves from the user's voice commands to obtain the original audio signal, and to perform noise reduction filtering on the original audio signal to obtain a clear voice segment.

[0040] The segmentation module is used to convert the clear audio segments into control text content, and to segment the control text content into keywords to obtain device identification words and operation instruction words;

[0041] The generation module is used to look up the corresponding target device communication address from the local device mapping table based on the device identifier word, and generate a structured control command frame based on the operation instruction word.

[0042] The sending module is used to perform signal encoding and modulation on the structured control command frame to obtain a radio frequency control signal, and send the radio frequency control signal to the home gateway. The home gateway then sends and controls the target smart device to perform a preset function action that matches the operation command word based on the target device's communication address.

[0043] This invention provides a control method for smart home devices, comprising the following steps: capturing sound waves from a user's voice command to obtain an original audio signal, and performing noise reduction filtering on the original audio signal to obtain a clear voice segment; converting the clear voice segment into control text content, and performing keyword segmentation on the control text content to obtain device identification words and operation command words; based on the device identification words, searching for the corresponding target device communication address from a local device mapping table, and generating a structured control command frame based on the operation command words; performing signal encoding and modulation on the structured control command frame to obtain a radio frequency control signal, and then... Radio frequency control signals are sent to the home gateway, which then sends and controls the target smart device to perform preset functional actions that match the operation command words based on the target device's communication address. This solves the technical problem that systems often struggle to accurately segment keywords and correctly match target devices with operation actions, leading to miscontrol or missed control. It achieves the extraction of device identifiers and operation command words through keyword segmentation technology, combined with a local device mapping table for rapid matching, avoiding reliance on cloud-based natural language understanding services. This not only reduces response latency but also improves the privacy and security of data processing, while supporting the technical effect of accurate parsing of multiple devices and multiple intent commands. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a schematic diagram of the steps of a control method for a smart home device in one embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of the steps of a control system for a smart home device in one embodiment of the present invention;

[0047] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0048] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0049] The following describes in detail, with reference to the accompanying drawings, a control method for a smart home device according to an embodiment of the present invention. First, the control method for a smart home device according to an embodiment of the present invention will be described with reference to the accompanying drawings.

[0050] Figure 1 This invention provides a method for controlling a smart home device, comprising the following steps:

[0051] Step S1: Capture the sound waves of the voice commands issued by the user to obtain the original audio signal, and perform noise reduction filtering on the original audio signal to obtain a clear voice segment.

[0052] Specifically, sound wave capture of user voice commands involves real-time acquisition of sound wave signals from the surrounding space using microphone arrays deployed in the smart home environment. This yields the raw audio signal containing the voice information. This raw audio signal is often mixed with environmental noise, appliance operation sounds, or human voice interference, thus requiring noise reduction and filtering. The process first converts the raw audio signal into a digital signal and inputs it into a digital signal processor. Then, an adaptive filtering algorithm combined with a preset noise model is used to suppress non-voice segments in the signal, especially filtering out steady-state noises common in home environments such as air conditioner operation sounds and television background noise in the frequency domain. Simultaneously, speech activity detection (VAD) technology is used to accurately identify speech segments and retain effective speech intervals. Subsequently, a bandpass filter is used to further enhance the signal strength in the human voice frequency range (usually 300Hz to 3400Hz), ultimately outputting a clear speech segment for subsequent accurate speech recognition and semantic analysis. For example, in a living room scenario, when a user says "turn on the living room air conditioner," the raw audio signal captured by the microphone may include the sound of the TV playing and the sound of traffic outside the window. After noise reduction filtering, the system can effectively reduce these interfering components and highlight the clear voice segment "turn on the living room air conditioner," providing high-quality audio input for subsequent conversion into control text content and ensuring the accurate start of the entire control process.

[0053] Step S2: Convert the clear audio segment into control text content, and perform keyword segmentation on the control text content to obtain device identification words and operation instruction words.

[0054] Specifically, the clear speech segment is input into a locally deployed speech recognition model. This model analyzes the audio frame sequence frame by frame based on a deep neural network and maps it to the corresponding Chinese characters or pinyin sequence, thereby generating control text content that matches the speech content. This control text content is a string in natural language form. Subsequently, word segmentation processing is performed on the control text content. A pre-trained Chinese word segmentation algorithm is used in conjunction with a smart home domain dictionary to segment the data, identifying device-specific words as device identifiers, and identifying verbs or phrases that indicate action intentions as operation command words. The entire process is completed in the local processor and does not rely on external network transmission. For example, in a living room scenario, when a user says "turn on the living room air conditioner," and after noise reduction processing to obtain a clear speech segment, the system uses speech recognition to convert it into the control text "turn on the living room air conditioner." Then, the system performs keyword segmentation on the text, extracting "living room air conditioner" as a device identifier and "turn on" as an operation command word. This allows the system to subsequently look up the corresponding target device communication address in the local device mapping table based on the device identifier and combine it with the operation command word to generate a structured control command frame. This enables precise control of the specified smart device and ensures that the semantic parsing process matches the actual device layout in the home.

[0055] Step S3: Based on the device identifier, find the corresponding target device communication address from the local device mapping table, and generate a structured control command frame based on the operation instruction word.

[0056] Specifically, based on the device identifier, a string matching search is performed in a local device mapping table pre-built and maintained in the local storage unit. This local device mapping table records the mapping relationship between the device identifier of each smart device in the home environment and its unique corresponding target device communication address. By performing precise or fuzzy matching between the extracted device identifier and the entries in the table, the associated target device communication address is determined. The target device communication address is a MAC address or IP address that conforms to the Zigbee or Wi-Fi communication protocol. At the same time, combined with the operation instruction words, a structured control command frame containing operation type, device address, execution parameters and check code fields is generated according to a predefined instruction encoding format. The structured control command frame is encapsulated in binary or JSON format to ensure that the syntax is standardized and can be parsed by the target device. For example, in a living room scenario, after keyword segmentation yields "living room air conditioner" as the device identifier and "turn on" as the operation command, the system looks up the target device communication address "192.168.1.105" corresponding to "living room air conditioner" in the local device mapping table. Based on the operation command "turn on", the system sets the operation type field of the control command frame to "POWER_ON" and the parameter field to the default operating mode. This ultimately generates a complete structured control command frame, providing standardized data input for subsequent signal encoding and modulation via the wireless communication module, ensuring accurate encapsulation and directional transmission of control commands.

[0057] Step S4: The structured control command frame is encoded and modulated to obtain a radio frequency control signal, which is then sent to the home gateway. The home gateway then sends and controls the target smart device to perform a preset function action that matches the operation command word based on the target device's communication address.

[0058] Specifically, the system receives structured control command frames from the local processor via a wireless communication module. This module has a built-in radio frequency chip compliant with the IEEE 802.11 or Zigbee protocol. First, the structured control command frame is digitally encoded using Manchester encoding or direct sequence spread spectrum technology to convert the binary data into a baseband signal suitable for wireless transmission. Then, the baseband signal is carrier-modulated using a modulation circuit, converting it into an airborne radio frequency control signal using FSK or QPSK modulation. This radio frequency control signal operates in the 2.4GHz band and maintains the same channel configuration as the home network. After modulation, the radio frequency control signal is transmitted via an antenna to the central node in the home network, i.e., the home gateway. The home gateway has wireless receiving capabilities and continuously listens to the designated communication channel. Upon receiving the radio frequency control signal, it demodulates and decodes it to restore the original structured control command frame, extracts the target device communication address, determines the forwarding path in the internal routing table based on the target device communication address, and finally forwards the control command to the corresponding target smart device. For example, in a living room scenario, when the system generates a structured control command frame containing the target device's communication address "192.168.1.105" and the operation command "POWER_ON", the wireless communication module encodes and modulates it into a 2.4GHz radio frequency control signal and sends it to the home gateway. After parsing the signal, the home gateway identifies the target device as an air conditioner connected to the same local area network, and sends a power-on command based on the IP address "192.168.1.105", causing the air conditioner to start the cooling mode and complete the preset function action matching the operation command word "turn on", realizing a complete closed-loop control process from voice input to device execution.

[0059] In a specific embodiment, the user's voice commands are captured to obtain the original audio signal, and the original audio signal is then subjected to noise reduction filtering to obtain a clear voice segment, including:

[0060] A sound wave sensor is used to capture the sound waves of the user's voice commands to obtain the raw audio signal, and the raw audio signal is subjected to spectrum analysis to obtain a spectrum distribution map;

[0061] Based on the spectrum distribution map, the original audio signal is subjected to noise frequency band identification to obtain the noise frequency band range. Then, the signal components in the original audio signal that are within the noise frequency band range are filtered out by the frequency band filter to obtain the preliminary noise-reduced audio signal.

[0062] The residual noise of the initial noise-reduced frequency signal is evaluated to obtain the residual noise intensity, and the gain of the initial noise-reduced frequency signal is adaptively adjusted based on the residual noise intensity to obtain a clear speech segment.

[0063] Specifically, the system uses a sound wave sensor to capture the sound waves from the user's voice commands. This is achieved by using a high-sensitivity microphone array deployed in the home environment to sense sound wave vibrations in the air in real time and convert them into continuous analog electrical signals. This results in a raw audio signal containing voice information and environmental interference. The raw audio signal is then sent to an analog-to-digital converter for sampling and quantization, forming a discrete digital signal that can be processed by a digital signal processor. Next, the raw audio signal is subjected to spectral analysis. The Fast Fourier Transform (FFT) algorithm is used to transform it from the time domain to the frequency domain, generating a spectral distribution map that reflects the energy distribution of each frequency component. This spectral distribution map visually displays the energy intensity of different frequency bands in the raw audio signal, making it easy to identify steady-state or periodic noise regions that are significantly higher than the background level. Based on the aforementioned spectrum distribution map, noise frequency bands are identified in the original audio signal. By setting an energy threshold and combining statistical methods, frequency ranges with sustained high energy and no speech characteristics are detected and determined as noise frequency bands. For example, the low-frequency humming of 500Hz to 800Hz generated by an air conditioner or the 600Hz narrowband interference caused by the start-stop of a refrigerator compressor can form obvious energy peak bands in the spectrum distribution map. After identification, the signal components in the original audio signal that fall within the noise frequency band range are filtered out by a frequency band filter. This frequency band filter adopts a digital notch filter or band-stop filter structure, and its parameters are dynamically configured according to the identified noise frequency band range to accurately attenuate the signal amplitude of the corresponding frequency band while retaining the speech information of the remaining frequency bands, thereby outputting a preliminary noise-reduced audio signal. The residual noise of the initial noise-reduced audio signal is evaluated by calculating parameters such as short-time energy, zero-crossing rate, and signal-to-noise ratio. The activity level of non-speech components still present in the audio stream is comprehensively judged to obtain the residual noise intensity. If the residual noise intensity exceeds a preset threshold, an adaptive gain adjustment mechanism is activated. This mechanism dynamically adjusts the gain coefficient of the speech band (usually 300Hz to 3400Hz) according to the residual noise intensity. While suppressing background fluctuations, it enhances the clarity of the human voice and avoids speech distortion caused by over-filtering. Finally, a clear speech segment with high signal-to-noise ratio and complete speech features is output.For example, in a living room scenario, when a user says "turn on the living room air conditioner," the raw audio signal captured by the sound wave sensor contains the dialogue from the TV and the low-frequency noise from the air conditioner itself. After spectrum analysis, a continuous high-energy band appears in the 400Hz–700Hz range. Based on this, the system identifies this range as a noise frequency band and filters it out using a band-stop filter to obtain a preliminary noise-reduced frequency signal. Subsequent evaluation reveals that there is still slight residual noise in the high-frequency band, with a moderate level of residual noise intensity. Therefore, a moderate gain boost is applied to the 1000Hz–2000Hz speech formant region to finally obtain a clear speech segment. This provides high-quality audio input for the subsequent accurate conversion into the control text content "turn on the living room air conditioner," ensuring that the entire voice control process can still start stably in complex acoustic environments.

[0064] In a specific embodiment, noise frequency bands are identified based on the spectrum distribution map of the original audio signal to obtain the noise frequency band range, including:

[0065] Energy peak detection is performed on the spectrum distribution map to obtain a set of energy peaks, and the energy peak set is sorted by amplitude to obtain a sorted peak list; based on the sorted peak list, the original audio signal is divided into frequency bands to obtain multiple candidate frequency bands, and the energy proportion of each candidate frequency band is calculated to obtain the energy proportion value of each frequency band; the energy proportion values ​​of each frequency band are filtered by a preset noise threshold to obtain the noise frequency band range.

[0066] Specifically, energy peak detection of the aforementioned spectrum distribution map involves traversing the amplitude values ​​of each frequency point within the frequency domain to identify discrete frequency points with significantly higher local energy than adjacent frequencies. These points constitute a set of energy peaks, each corresponding to a specific frequency and its energy intensity. The energy peak set is then sorted by amplitude, arranged in descending order of energy value, forming a sorted peak list. The peaks at the beginning of this list represent the frequency components with the most concentrated energy in the entire audio signal, typically closely related to persistent noise sources such as mechanical vibrations or electromagnetic interference from household appliances. Based on this sorted peak list, the original audio signal is divided into frequency bands. The top N significant peaks in the sorted peak list are used as the center frequency, and multiple candidate frequency bands are generated by extending a fixed bandwidth (e.g., ±50Hz) to the left and right. Each candidate frequency band covers a frequency range centered on a specific high-energy peak, ensuring that potential noise concentration areas are fully encompassed. This division method relies on actual spectral characteristics rather than fixed frequency band settings, improving the targeting of noise identification. The energy percentage calculation for each candidate frequency band involves integrating and summing the energy amplitudes of all frequency points within that band, then dividing by the total energy of the entire spectrum distribution. This yields the energy percentage of each band, reflecting its proportion of the overall signal. A higher value indicates a greater contribution of that band to the overall noise. The energy percentages of each band are then filtered using a preset noise threshold. This involves comparing the calculated energy percentages with a pre-defined noise threshold. If the energy percentage of a candidate band exceeds this threshold, it is included in the noise band range. This preset noise threshold is set based on a statistical model of home environmental noise, for example, 12% or 15%, ensuring that only strong interference bands significantly affecting speech clarity are retained, avoiding misjudgment of weak noise areas. For example, in a living room scenario, when a user says "turn on the living room air conditioner," the raw audio signal captured by the sound wave sensor is analyzed to generate a spectrum distribution map. The system detects significant energy peaks at 550Hz, 620Hz, and 800Hz, forming a set of energy peaks. After amplitude sorting, a sorted list of peaks is generated. Taking the first three peaks as the center, three candidate frequency bands are divided into 500–600Hz, 570–670Hz, and 750–850Hz, respectively. Their energy proportions are calculated to be 18%, 16%, and 9%, respectively. After filtering by a preset noise threshold of 15%, the first two candidate frequency bands are confirmed as noise frequency bands, namely 500–600Hz and 570–670Hz, because their energy proportions exceed the threshold. This result highly matches the typical low-frequency humming frequency band generated by the air conditioner compressor during operation. Subsequently, the system will perform filtering on these two frequency bands to effectively remove the device's own noise, providing accurate noise localization basis for obtaining high-quality, clear voice segments, and ensuring the accuracy and stability of voice command parsing.

[0067] In a specific embodiment, converting the clear speech segment into control text content includes:

[0068] Speech features are extracted from the clear speech segment to obtain a speech feature sequence, and acoustic unit matching is performed on the speech feature sequence to obtain an acoustic unit sequence.

[0069] Based on the acoustic unit sequence, a pronunciation mapping is performed to obtain a pinyin symbol string, and the pinyin symbol string is segmented into continuous syllables to obtain syllable unit groups;

[0070] The syllable unit group is matched with words to obtain a candidate word set, and the candidate word set is verified for contextual association to obtain the control text content.

[0071] Specifically, speech feature extraction of the clear speech segments involves segmenting the noise-reduced, filtered speech segments into frames, each typically 25 milliseconds long with a 10-millisecond frame shift. Mel-frequency cepstral coefficients (MFCCs) are calculated for each frame, extracting multidimensional features including energy, zero-crossing rate, and first and second-order differences to form a time-sequential speech feature sequence. This sequence characterizes the dynamic changes of the speech signal in the time-frequency domain. Acoustic unit matching of the speech feature sequence involves inputting it into a locally deployed acoustic model. This model, based on a Hidden Markov Model (HMM) or Deep Neural Network (DNN), is pre-trained with a large number of acoustic units for Chinese speech, such as phonemes or triphones. By calculating the likelihood probability between the speech feature sequence and each acoustic unit, the most matching path is found, outputting the corresponding acoustic unit sequence. This sequence consists of a series of phoneme symbols representing the basic units of speech pronunciation. The pronunciation mapping based on the acoustic unit sequence involves converting the acoustic unit sequence according to the rules of Chinese Pinyin. For example, the initial consonant "zh", the final vowel "ong", and the tone "2" are mapped to "zhong²". The entire acoustic unit sequence is converted into a continuous string of Pinyin symbols through table lookup and rule derivation, which retains the pronunciation information of the original speech. The Pinyin symbol string is then segmented into continuous syllables. Based on the syllable boundary rules of Chinese speech, combined with silence detection and energy mutation point identification, segmentation points are inserted between syllables. For example, "dakai ketiting de kongtiao" is segmented into "da-kai-ke-ting-de-kong-tiao", forming syllable unit groups composed of single syllables, each syllable corresponding to the pronunciation of a Chinese character. Lexical matching of the syllable unit groups involves searching the syllable unit groups in the vocabulary of the local language model and using N-gram or dictionary-based matching algorithms to find possible word combinations. For example, "da kai" matches "open", "ke ting" matches "living room", and "kong tiao" matches "air conditioner", generating a candidate word set containing multiple possible paths. Contextual correlation verification of the candidate word set involves combining contextual rules in the smart home field and commonly used user command patterns to evaluate the semantic rationality of each candidate path. For example, paths containing device names and verb combinations are prioritized, while fragments such as "da k" (check-in) and "living room" (living room) that do not conform to control logic are excluded. Finally, the text sequence that best fits the context is determined as the control text content.For example, in a living room scenario, when a user says "turn on the living room air conditioner," the system extracts a speech feature sequence and obtains an acoustic unit sequence of "dak-ai-ket-ing-dek-ong-t-iao" through acoustic unit matching. This sequence is then mapped to the pinyin string "da kai ke ting dekong tiao" through pronunciation mapping. Continuous syllable segmentation is then used to obtain syllable unit groups. Word matching generates candidate words such as "open / check in," "living room," "of," and "air conditioner." After contextual correlation verification, irrelevant words such as "check in" are excluded, and "turn on the living room air conditioner" is retained as the final control text content. This provides accurate natural language input for subsequent keyword segmentation, ensuring the correctness of command parsing and scenario adaptability.

[0072] In a specific embodiment, the control text content is segmented by keywords to obtain device identification words and operation instruction words, including:

[0073] The control text content is initially segmented into character blocks to obtain a set of character blocks. Part-of-speech tagging is then performed on the character block set to obtain a part-of-speech tag set. Based on the part-of-speech tag set, the character block set is further divided into semantic units to obtain semantic unit groups. Device lexicon matching is performed on the semantic unit groups to obtain a set of candidate device words. Contextual association verification is then performed on the candidate device word set to obtain device identifier words. After removing word blocks containing device identifier words, operation instruction lexicon matching is performed on the remaining semantic unit groups in the semantic unit groups to obtain a set of candidate operation words. Functional logic verification is then performed based on the device identifier words and the set of candidate operation words to obtain operation instruction words.

[0074] Specifically, for the preliminary character segmentation of the control text content, the control text content generated through speech recognition is segmented character by character according to individual Chinese characters or Chinese punctuation marks, forming a character block set composed of independent characters. This character block set retains the linear sequence structure of the original text. Subsequently, for the词性标注 of the character block set, using a locally preset Chinese word segmentation and词性标注 model, combined with a conditional random field (CRF) or a BiLSTM network, corresponding词性标签s such as verbs, nouns, auxiliary words, adjectives, etc. are assigned to each character or continuous character combination, generating a词性标记 set corresponding to the order of the character block set. This词性标记 set provides a linguistic basis for subsequent syntactic structure analysis. Based on the词性标记 set, for the semantic unit division of the character block set, according to the rules of Chinese grammar structure, words or phrases with complete semantic functions are combined and extracted from the character sequence. For example, "open" (verb), "in the living room" (attributive), and "air conditioner" (noun) are respectively divided into independent semantic units, forming a semantic unit group. This semantic unit group is composed of multiple semantic segments with clear grammatical roles, facilitating subsequent targeted matching and verification. For the device vocabulary matching of the semantic unit group, each semantic unit in the semantic unit group is compared with the strings in the locally stored device vocabulary. This device vocabulary pre-enters the registered names and their common aliases of all smart devices in the home, such as "air conditioner in the living room", "master bedroom light", "kitchen socket", etc. Through fuzzy matching or exact matching, semantic units that match the device name are identified, forming a candidate device word set. For example, "air conditioner", "living room air conditioner", and "air conditioner in the living room" may all be identified as candidates. For the context association verification of the candidate device word set, logical judgments are made by combining the positional relationship, modifying components, and the preceding and following词性 in the semantic unit group. Candidates that contain a spatial qualifier (such as "in the living room") and have a complete syntactic structure (such as a modifier-head structure) are preferred, and isolated or incomplete expressions are excluded. For example, in "open the air conditioner in the living room", "air conditioner in the living room" is confirmed as a valid device identifier because it contains a clear spatial attribute and has a complete structure with the word "的", while "air conditioner" is excluded due to the lack of positioning information. Finally, "air conditioner in the living room" is determined as the device identifier. After removing the word blocks containing the device identifier, for the operation instruction vocabulary matching of the remaining semantic unit group in the semantic unit group, the remaining part such as "open" is sent to the operation instruction vocabulary for comparison. This vocabulary contains standard operation verbs such as "open", "close", "raise", "start", "pause", etc. Through matching, a candidate operation word set is obtained. For example, "open" is identified as a candidate. For the functional logic verification based on the device identifier and the candidate operation word set, the function capability table of the device corresponding to the device identifier is queried. For example, "air conditioner in the living room" supports operations such as "turn on", "turn off", "adjust temperature", etc., and "open" belongs to its legal operation instruction. Therefore, through verification, "open" is finally confirmed as the operation instruction word. It should be noted that there seem to be some inaccuracies or unclear parts in the original text where the same term "词性标注" is repeated and not fully explained in a consistent way. The above translation tries to make sense of the overall context as accurately as possible while keeping the original text structure.For example, in a living room scenario, the control text content is "turn on the living room air conditioner". After initial character segmentation, character blocks such as "open / open / living / room / air / conditioner" are obtained. After part-of-speech tagging, verb, noun, and auxiliary word tags are generated. The semantic units are divided into "open", "living room", and "air conditioner", which are combined into two semantic units: "open" and "living room air conditioner". After matching with the device lexicon, "living room air conditioner" enters the candidate device word set. Contextual verification confirms that it is a complete device name with clear spatial location, becoming a device identifier word. After removal, the remaining "open" is successfully matched in the operation instruction lexicon. Combined with the functional logic of "air conditioner" supporting the "open" operation, it passes the verification and is finally determined to be the operation instruction word, providing accurate input for subsequent search of target device communication address and generation of structured control command frames.

[0075] In a specific embodiment, based on the device identifier, the corresponding target device communication address is looked up from the local device mapping table, including:

[0076] The local device mapping table is parsed to obtain a device information list, and the device identifier field and communication address field in the device information list are associated and marked to obtain an association mapping table;

[0077] The device identifier is processed by character standardization to obtain a standard identifier, and a field matching query is performed on the association mapping table based on the standard identifier to obtain candidate communication addresses;

[0078] The candidate communication addresses are counted to obtain the address count result, and the uniqueness of the candidate communication addresses is verified based on the address count result to obtain the target device communication address.

[0079] Specifically, parsing the local device mapping table involves reading the table, stored in local non-volatile memory, in a structured data format. This local device mapping table is typically organized as a JSON, XML, or SQLite database and contains multiple record entries, each corresponding to a registered smart device. The field parsing process reads the table line by line, identifying preset field names such as the "Device Identifier Field" and the "Communication Address Field." The Device Identifier Field stores the device's name or alias, such as "Living Room Air Conditioner" or "Master Bedroom Light." The Communication Address Field stores the unique identifier of the corresponding device on the network, such as an IP address, MAC address, or Zigbee short address. These field contents are extracted through parsing operations to form a structured device information list. This device information list is a two-dimensional data structure, with each row representing a device record and each column corresponding to a field value. Associating the device identifier field and communication address field in the device information list involves establishing a one-to-one correspondence between the device identifier field value and its corresponding communication address field value in the same record row, organized in key-value pairs. For example, "living room air conditioner" is used as the key, and "192.168.1.105" is used as the value. Multiple key-value pairs together form an association mapping table, which supports fast lookup operations and improves subsequent matching efficiency. Standardizing the device identifier words involves normalizing possible differences in expression after user speech recognition. For example, "living room air conditioner," "living room air conditioner," and "living room air conditioner" are uniformly converted to the standard naming format "living room air conditioner." The process includes removing redundant auxiliary words, completing omitted structures, correcting homophones and typos, and standardizing quantifier expressions to ensure consistency with the device identifier field format registered in the local device mapping table, resulting in standard identifier words. The field matching query based on the standard identifier in the association mapping table involves using the standard identifier as the query keyword to perform string matching across all keys in the association mapping table. Both exact matching and fuzzy matching modes are supported. If a completely identical key exists, the corresponding communication address is directly returned as a candidate communication address. If no exact match is found, a fuzzy matching algorithm, such as edit distance or pinyin similarity calculation, is used to find the closest match and add its communication address to the candidate communication address set. A count of the candidate communication addresses is performed, which involves counting the total number of communication addresses returned in the field matching query results. This count is used to determine if there are any anomalies such as multiple device matches or no matches.Based on the address count result, the uniqueness of the candidate communication address is verified. If the address count result is 1, the candidate communication address is confirmed as a valid and unique target device communication address and can be directly used for subsequent control command generation. If the address count result is greater than 1, it indicates that there are multiple devices with similar names or registration conflicts. The system will trigger a conflict handling mechanism and further filter based on contextual information such as room location, device type, or user's historical operation preferences. If the address count result is 0, it is determined that the device is not registered or is incorrectly identified, an error code is returned, and the control process is terminated. For example, in a living room scenario, after keyword segmentation yields "living room air conditioner" as a device identifier, the system parses the local device mapping table to extract a list of device information containing entries such as "living room air conditioner: 192.168.1.105" and "kitchen light: 192.168.1.106". It then establishes an association mapping table between the device identifier field and the communication address field. Subsequently, "living room air conditioner" undergoes character standardization to ensure its format matches the registered name. This is then used as a standard identifier in the association mapping table, returning the unique candidate communication address "192.168.1.105". After counting, the number of addresses is 1. This uniqueness check confirms that this address is the final target device communication address, providing accurate routing information for generating structured control command frames and ensuring that control commands are accurately delivered to the designated device.

[0080] In a specific embodiment, the structured control command frame is signal encoded and modulated to obtain a radio frequency control signal, including:

[0081] The structured control command frame is parsed to obtain a digital instruction sequence, and the frame header and frame tail of the digital instruction sequence are checked to obtain the checked digital instruction.

[0082] Based on the wireless communication protocol, the verified digital instructions are subjected to symbol mapping conversion to obtain a set of modulation symbols, and the set of modulation symbols is optimized by constellation diagram distribution to obtain optimized modulation symbols;

[0083] The optimized modulation symbol is modulated by carrier frequency using radio frequency modulation technology to obtain a preliminary radio frequency control signal. The power spectral density of the radio frequency control signal is then detected to obtain a radio frequency control signal that conforms to the transmission standard.

[0084] Specifically, data format parsing of the structured control command frame involves disassembling the structured control command frame, generated from device identifiers and operation instruction words, according to a predefined communication protocol format. This structured control command frame is typically encapsulated in binary or compact text format and includes components such as a frame header, device address field, opcode field, parameter field, checksum field, and frame trailer. The data format parsing process first identifies the frame header flag to determine the start position of the frame, and then extracts the target device communication address, operation instruction code, and additional parameters sequentially according to a fixed offset to form a complete digital instruction sequence. This digital instruction sequence represents all the information of the control action in binary stream form. The frame header and trailer verification of the digital instruction sequence is performed by verifying whether the fixed identifier codes of the frame header and trailer conform to the protocol specifications. For example, the frame header is 0xAA55 and the frame trailer is 0x55AA. If the actual read value is consistent with the preset value, the frame boundary is determined to be correct. At the same time, cyclic redundancy check (CRC) or checksum calculation is performed on the digital instruction sequence, and the calculation result is compared with the check field carried in the frame. If they are consistent, the data integrity is confirmed to be correct, and the verified digital instruction is obtained. If they are inconsistent, the frame is discarded and the retransmission mechanism is triggered to ensure the reliability of the transmitted data. Based on wireless communication protocols, the verified digital instructions undergo symbol mapping conversion. This is done according to the physical layer specifications of current wireless communication standards such as IEEE 802.11n or Zigbee 3.0. The binary digital instructions are divided into several bit groups according to the modulation order. For example, in QPSK modulation, every 2 bits are mapped to one symbol, and in 16-QAM, every 4 bits are mapped to one symbol. Each bit group is converted into a corresponding complex form modulation symbol using a lookup table, forming a set of modulation symbols. This set represents the discrete representation of digital information in the complex plane. Constellation diagram distribution optimization of the modulation symbol set involves introducing pre-distortion compensation and power normalization algorithms during modulation to adjust the position distribution of each modulation symbol in the constellation diagram. This reduces the Euclidean distance deviation between symbols, suppresses constellation point spread caused by power amplifier nonlinearity or phase noise, and improves the signal's anti-interference capability during transmission, resulting in optimized modulation symbols. This optimization process ensures that the receiving end can more accurately demodulate and restore the original data. The optimized modulation symbol is carrier frequency modulated by radio frequency modulation technology. The optimized modulation symbol is input to a quadrature modulator and multiplied with the in-phase (I) and quadrature (Q) carrier signals generated by the local oscillator to synthesize an intermediate frequency analog signal. Then, the signal is shifted to the 2.4 GHz or 5.8 GHz radio frequency band by an up-conversion circuit to form a preliminary radio frequency control signal. This signal already has the physical characteristics to propagate in a wireless channel.The power spectral density detection of the radio frequency control signal involves using a spectrum analysis module to monitor the frequency domain energy distribution of the signal in real time. This ensures that the power spectral density complies with the limits set by national radio management agencies, such as the FCC or SRRC for the transmit power of smart home devices. If an energy limit is detected in a certain frequency band, the transmit power is dynamically adjusted through an automatic gain control (AGC) circuit to make the overall power spectrum distribution of the signal smooth and compliant, ultimately outputting a radio frequency control signal that meets transmission standards. For example, in a living room scenario, when a structured control command frame contains the target device communication address "192.168.1.105" and the opcode "POWER_ON", the system parses it into a digital instruction sequence "AA55 10500001 ON 0F 55AA". After verifying that the frame header and tail are correct, the system uses QPSK modulation according to the Zigbee protocol to map it into a set of modulation symbols. Then, it optimizes the constellation diagram to reduce inter-symbol interference, generates a preliminary 2.4GHz radio frequency control signal through radio frequency modulation, and finally ensures that its transmission intensity is within the legal range through power spectral density detection, thereby obtaining a standard-compliant radio frequency control signal, which provides physical layer support for subsequent transmission to the home gateway.

[0085] In a specific embodiment, constellation diagram distribution optimization is performed on the modulation symbol set to obtain optimized modulation symbols, including:

[0086] The initial constellation points are located in the set of modulation symbols to obtain an initial constellation point position set, and the distance between neighboring points is calculated in the initial constellation point position set to obtain a neighboring point distance set.

[0087] Based on the set of distances to neighboring points, the initial set of constellation point positions is filtered for outliers to obtain anomaly constellation point summation. The positions of the outlier constellation points are then corrected by moving them toward the center of their neighboring normal constellation points to obtain the corrected constellation point positions.

[0088] The modulation symbol set is remapped based on the corrected constellation point positions to obtain optimized modulation symbols.

[0089] Specifically, the initial constellation point localization of the modulation symbol set involves representing each complex symbol in the modulation symbol set, after symbol mapping transformation, with coordinates on a two-dimensional complex plane. This complex plane uses the real part as the horizontal axis and the imaginary part as the vertical axis, with each modulation symbol corresponding to a specific coordinate point, forming the initial constellation point position set. For example, in QPSK modulation, ideally there should be four equally spaced constellation points located near (1,1), (1,-1), (-1,1), and (-1,-1). However, due to nonlinear distortion, phase jitter, or quantization errors during signal processing, the actual constellation points may deviate from the ideal positions. The neighbor distance calculation of the initial constellation point position set involves calculating the Euclidean distance between each initial constellation point and other surrounding constellation points, forming a neighbor distance set. This set records the actual distance between all adjacent constellation point pairs and is used to evaluate the uniformity of the constellation diagram distribution. If the distance between a point and its neighboring points is significantly less than the average distance, it may cause demodulation misjudgment. The initial constellation point location set is screened for anomalies based on the neighboring point distance set. This involves setting a distance threshold range, which is based on the minimum theoretical spacing under the ideal modulation mode and introduces a safety margin. Constellation points involved in point pairs in the neighboring point distance set that are lower than the threshold are marked as potential anomalies. Combined with their topological position and frequency of occurrence in the constellation diagram, abnormal constellation points that deviate from the normal distribution pattern are identified. For example, a symbol that should be in the (1,1) region frequently appears in the (0.3,0.4) position and is too close to neighboring points, and is then determined to be an abnormal constellation point, thus forming an abnormal constellation point combination. The position correction of the abnormal constellation points involves analyzing the spatial distribution of nearby normal constellation points that are not marked as abnormal around each identified abnormal constellation point, and calculating the geometric center position of these normal points. For example, if there are three normal points around an abnormal point located at (1,1), (1,-1), and (-1,1), then its center position is (0.33, 0.33). The coordinates of the abnormal constellation point are then appropriately shifted towards this center position. The shift magnitude is weighted and adjusted according to the original deviation to avoid overcorrection leading to new distortions. The corrected constellation point positions are then obtained. This process effectively alleviates the problem of constellation diagram shrinkage or diffusion caused by non-ideal hardware characteristics. Based on the corrected constellation point positions, the modulation symbol set is remapped. This involves replacing the symbols in the original modulation symbol set corresponding to the abnormal constellation points with their corrected coordinate values, making the constellation distribution of the entire modulation symbol set more regular and the isolation between points higher, thereby forming optimized modulation symbols and improving the signal's noise immunity and demodulation accuracy in the wireless channel.For example, in a living room scenario, after the system generates a structured control command frame to control the "living room air conditioner" and converts it into QPSK modulation symbols, the initial constellation point positioning finds that some symbols are concentrated in the (0.5, 0.5) region. The distance calculation of neighboring points shows that the distance from the ideal (1, 1) point is only 0.7, which is lower than the set threshold of 1.0. After screening out anomaly points, this batch of symbols is confirmed to be an abnormal constellation point. The system calculates that the center position of the normal points (1, -1), (-1, 1), and (-1, -1) around it is approximately (-0.33, -0.33), and moves the abnormal point in this direction to (0.8, 0.8) to complete the position correction. Then, based on the corrected constellation point position, the original symbols are remapped, and optimized modulation symbols are output to ensure that the signal after radio frequency modulation has higher transmission reliability, reduce the risk of bit errors caused by multipath effects or electrical interference in the home environment, and ensure that the control command is accurately delivered to the target device.

[0090] In a specific embodiment, the radio frequency control signal is sent to the home gateway, and the home gateway, based on the target device's communication address, sends and controls the target smart device to execute a preset function action matching the operation command word, including:

[0091] The radio frequency control signal is encapsulated with a protocol header to obtain a radio frequency signal with a protocol header; an address field is embedded in the radio frequency signal with the protocol header based on the communication address of the target device to obtain an address-bearing radio frequency signal; the address-bearing radio frequency signal is sent to the home gateway, and a routing path is planned for the address-bearing radio frequency signal based on the routing table of the home gateway to obtain planned path information; the address-bearing radio frequency signal is sent to the target smart device through the wireless transmission module of the home gateway according to the planned path information;

[0092] The address-bearing radio frequency signal is frame-parsed based on the protocol parsing engine technology in the target smart device to extract the operation instruction words in the address-bearing radio frequency signal, and the operation instruction words are encoded and converted based on the function mapping table stored in the home gateway to obtain the preset function code.

[0093] The preset function code is sent to the target smart device corresponding to the device matching result, and the target smart device is controlled to perform the action corresponding to the preset function code.

[0094] Specifically, the protocol header encapsulation of the radio frequency control signal involves adding a control information header, as specified by the communication protocol, to the already modulated and optimized radio frequency control signal that conforms to transmission standards during the baseband processing stage. This protocol header includes metadata fields such as version number, data type identifier, hop count limit, and quality of service level, which guide the correct processing at the network and link layers. The encapsulation process is completed in the communication protocol stack at the transmitting end, ensuring that the radio frequency signal has a complete protocol structure, forming a radio frequency signal with a protocol header. This signal not only carries the original control command content but also contains the routing and management information required for network transmission. Address field embedding of the radio frequency signal with the protocol header based on the target device's communication address involves writing the target device's communication address found in the previous steps into the address field of the protocol header as a destination address. This address field is usually located at a fixed offset position in the protocol header and is represented using a standard encoding format such as IPv4 address or MAC address. The embedding operation ensures that the signal can be recognized and correctly forwarded by intermediate nodes during transmission, ultimately generating a radio frequency signal with an address. This signal has a clear destination identifier, providing a basis for subsequent routing decisions. The addressed radio frequency signal is sent to the home gateway by the radio frequency front-end circuit of the wireless communication module driving the antenna to transmit modulated high-frequency electromagnetic waves into the air. The wireless receiving unit built into the home gateway continuously listens on the designated channel, captures the signal, demodulates and decodes it to restore the digital form of the addressed radio frequency signal, and then transmits it to the data processing unit inside the gateway. Based on the routing table of the home gateway, a routing path is planned for the addressed radio frequency signal. The home gateway parses the target device communication address in the protocol header and searches for a matching forwarding entry in the locally maintained routing table. This routing table records the mapping relationship between the communication addresses of all smart devices in the home network and the physical ports or wireless channels they are connected to. The access path of the next hop node or the direct target device is determined by the longest prefix matching or exact matching algorithm, generating planned path information. This information includes control parameters such as output interface, forwarding priority, and transmission protocol type. The home gateway's wireless transmission module transmits the addressed radio frequency signal to the target smart device according to the planned path information. The home gateway configures the operating parameters of its wireless transmission module, such as transmission power, channel number, and modulation method, based on the planned path information. The signal is then directionally transmitted to the network node where the target device is located via short-range wireless technologies such as Wi-Fi, Zigbee, or Bluetooth. If the target device is in a sleep state, the gateway can trigger a wake-up mechanism to ensure reliable signal delivery.The target smart device uses a protocol parsing engine to perform frame parsing on the address-bearing radio frequency signal. Upon receiving the signal, the target device activates its built-in protocol parsing engine. This engine peels off the protocol header layer by layer according to a preset communication protocol specification, extracts the payload, and reconstructs the original operation command words, such as "open" or "POWER_ON". This process includes verification, decryption (e.g., enabling secure communication), and field extraction to ensure the integrity of the command content. The operation command words are then encoded and converted based on a function mapping table stored in the home gateway. This allows the home gateway to convert the operation command words into device-specific preset function codes before or during forwarding, based on its locally stored function mapping table. This function mapping table records the correspondence between general command terms and specific device control codes; for example, "open" is mapped to "0x01" and "close" to "0x00". This conversion process supports command compatibility across multiple brands of devices, avoiding control failures due to differences in proprietary protocols among device manufacturers. Sending the preset function code to the target smart device corresponding to the device matching result involves the home gateway encapsulating the converted preset function code into a control message of a device-specific format through the established communication link and sending it. After receiving the code, the target smart device triggers its internal execution unit to drive the hardware module to perform the corresponding action. For example, in a living room scenario, when a user issues the voice command "turn on the living room air conditioner," the system generates a structured control command frame with the target device communication address "192.168.1.105." After being modulated into a radio frequency control signal, a protocol header is added and the address is embedded to form an address-enabled radio frequency signal, which is then sent to the home gateway. The gateway determines a terminal in the Wi-Fi subnet corresponding to this address based on its routing table, plans the path information to the 2.4GHz channel 6 of the wireless transmission module, and sends the signal to the air conditioner through this module. The air conditioner's built-in protocol parsing engine parses the "turn on" command, and at the same time, the home gateway queries the function mapping table to convert "turn on" into the preset function code "0x01" and sends it to the air conditioner. Finally, the air conditioner is controlled to start the cooling mode, completing the preset function action matching the operation command, realizing a closed-loop control process from voice input to device execution.

[0095] The control method for smart home devices in embodiments of the present invention has been described above. The control system for smart home devices in embodiments of the present invention is described below. Please refer to [link / reference]. Figure 2 One embodiment of the control system for smart home devices in this invention includes:

[0096] The capture module 21 is used to capture sound waves from the user's voice commands to obtain the original audio signal, and to perform noise reduction filtering on the original audio signal to obtain a clear voice segment.

[0097] The segmentation module 22 is used to convert the clear speech segment into control text content, and to perform keyword segmentation on the control text content to obtain device identification words and operation instruction words;

[0098] The generation module 23 is used to look up the corresponding target device communication address from the local device mapping table based on the device identifier word, and generate a structured control command frame based on the operation instruction word;

[0099] The sending module 24 is used to perform signal encoding and modulation on the structured control command frame to obtain a radio frequency control signal, and send the radio frequency control signal to the home gateway. The home gateway then sends and controls the target smart device to perform a preset function action that matches the operation command word based on the target device's communication address.

[0100] In this embodiment, the specific implementation of each module in the above system embodiment is described in the above method embodiment, and will not be repeated here.

Claims

1. A control method for a smart home device, characterized in that, Includes the following steps: The user's voice commands are captured to obtain the original audio signal, and the original audio signal is then subjected to noise reduction filtering to obtain a clear voice segment. The clear audio segment is converted into control text content, and the control text content is segmented by keywords to obtain device identification words and operation instruction words; Based on the device identifier, the corresponding target device communication address is found in the local device mapping table, and a structured control command frame is generated based on the operation instruction word; The structured control command frame is encoded and modulated to obtain a radio frequency control signal, which is then sent to the home gateway. The home gateway, based on the target device's communication address, sends and controls the target smart device to perform a preset function action that matches the operation command word.

2. The control method for smart home devices according to claim 1, characterized in that, The system captures sound waves from user-issued voice commands to obtain the original audio signal, and then performs noise reduction filtering on the original audio signal to obtain clear speech segments, including: A sound wave sensor is used to capture the sound waves of the user's voice commands to obtain the raw audio signal, and the raw audio signal is subjected to spectrum analysis to obtain a spectrum distribution map; Based on the spectrum distribution map, the original audio signal is subjected to noise frequency band identification to obtain the noise frequency band range. Then, the signal components in the original audio signal that are within the noise frequency band range are filtered out by the frequency band filter to obtain the preliminary noise-reduced audio signal. The residual noise of the initial noise-reduced frequency signal is evaluated to obtain the residual noise intensity, and the gain of the initial noise-reduced frequency signal is adaptively adjusted based on the residual noise intensity to obtain a clear speech segment.

3. The control method for smart home devices according to claim 2, characterized in that, Based on the aforementioned spectral distribution map, noise frequency bands are identified in the original audio signal to obtain the noise frequency band range, including: Energy peak detection is performed on the spectrum distribution map to obtain a set of energy peaks, and the energy peak set is sorted by amplitude to obtain a sorted list of peaks; Based on the sorted peak list, the original audio signal is divided into frequency bands to obtain multiple candidate frequency bands, and the energy proportion of each candidate frequency band is calculated to obtain the energy proportion value of each frequency band. The noise frequency range is obtained by filtering the energy proportion values ​​of each frequency band by setting a preset noise threshold.

4. The control method for smart home devices according to claim 1, characterized in that, Converting the clear speech segment into control text content includes: Speech features are extracted from the clear speech segment to obtain a speech feature sequence, and acoustic unit matching is performed on the speech feature sequence to obtain an acoustic unit sequence. Based on the acoustic unit sequence, a pronunciation mapping is performed to obtain a pinyin symbol string, and the pinyin symbol string is segmented into continuous syllables to obtain syllable unit groups; The syllable unit group is matched with words to obtain a candidate word set, and the candidate word set is verified for contextual association to obtain the control text content.

5. The control method for smart home devices according to claim 1, characterized in that, The control text content is segmented by keywords to obtain device identification words and operation instruction words, including: The control text content is initially segmented into characters to obtain a set of character blocks. The set of character blocks is then labeled with parts of speech to obtain a set of part-of-speech tags. Based on the set of part-of-speech tags, the set of character blocks is divided into semantic units to obtain semantic unit groups. The semantic unit group is matched with a device lexicon to obtain a candidate device word set, and the candidate device word set is verified by context association to obtain device identifier words; After removing word blocks containing device identifiers, the remaining semantic unit groups in the semantic unit group are matched with the operation instruction word library to obtain a set of candidate operation words. Then, based on the device identifier and the set of candidate operation words, functional logic verification is performed to obtain the operation instruction words.

6. The control method for smart home devices according to claim 1, characterized in that, Based on the device identifier, the corresponding target device communication address is searched from the local device mapping table, including: The local device mapping table is parsed to obtain a device information list, and the device identifier field and communication address field in the device information list are associated and marked to obtain an association mapping table; The device identifier is processed by character standardization to obtain a standard identifier, and a field matching query is performed on the association mapping table based on the standard identifier to obtain candidate communication addresses; The candidate communication addresses are counted to obtain the address count result, and the uniqueness of the candidate communication addresses is verified based on the address count result to obtain the target device communication address.

7. The control method for smart home devices according to claim 1, characterized in that, The structured control command frame is signal encoded and modulated to obtain a radio frequency control signal, including: The structured control command frame is parsed to obtain a digital instruction sequence, and the frame header and frame tail of the digital instruction sequence are checked to obtain the checked digital instruction. Based on the wireless communication protocol, the verified digital instructions are subjected to symbol mapping conversion to obtain a set of modulation symbols, and the set of modulation symbols is optimized by constellation diagram distribution to obtain optimized modulation symbols; The optimized modulation symbol is modulated by carrier frequency using radio frequency modulation technology to obtain a preliminary radio frequency control signal. The power spectral density of the radio frequency control signal is then detected to obtain a radio frequency control signal that conforms to the transmission standard.

8. The control method for smart home devices according to claim 7, characterized in that, The set of modulation symbols is optimized by constellation diagram distribution to obtain optimized modulation symbols, including: The initial constellation points are located in the set of modulation symbols to obtain an initial constellation point position set, and the distance between neighboring points is calculated in the initial constellation point position set to obtain a neighboring point distance set. Based on the set of distances to neighboring points, the initial set of constellation point positions is filtered for outliers to obtain anomaly constellation point summation. The positions of the outlier constellation points are then corrected by moving them toward the center of their neighboring normal constellation points to obtain the corrected constellation point positions. The modulation symbol set is remapped based on the corrected constellation point positions to obtain optimized modulation symbols.

9. The control method for smart home devices according to claim 1, characterized in that, The radio frequency control signal is sent to the home gateway, and the home gateway, based on the target device's communication address, sends and controls the target smart device to execute a preset function action matching the operation command word, including: The radio frequency control signal is encapsulated with a protocol header to obtain a radio frequency signal with a protocol header; The address field of the radio frequency signal with protocol header is embedded based on the communication address of the target device to obtain the radio frequency signal with address. The radio frequency signal with address is sent to the home gateway, and the routing path of the radio frequency signal with address is planned based on the routing table of the home gateway to obtain the planned path information. The address-bearing radio frequency signal is transmitted to the target smart device through the wireless transmission module of the home gateway, according to the planned path information. The address-bearing radio frequency signal is frame-parsed based on the protocol parsing engine technology in the target smart device to extract the operation instruction words in the address-bearing radio frequency signal, and the operation instruction words are encoded and converted based on the function mapping table stored in the home gateway to obtain the preset function code. The preset function code is sent to the target smart device corresponding to the device matching result, and the target smart device is controlled to perform the action corresponding to the preset function code.

10. A smart home system for performing a control method for a smart home device as described in any one of claims 1-9, comprising: The capture module is used to capture sound waves from the user's voice commands to obtain the original audio signal, and to perform noise reduction filtering on the original audio signal to obtain a clear voice segment. The segmentation module is used to convert the clear audio segments into control text content, and to segment the control text content into keywords to obtain device identification words and operation instruction words; The generation module is used to look up the corresponding target device communication address from the local device mapping table based on the device identifier word, and generate a structured control command frame based on the operation instruction word. The sending module is used to perform signal encoding and modulation on the structured control command frame to obtain a radio frequency control signal, and send the radio frequency control signal to the home gateway. The home gateway then sends and controls the target smart device to perform a preset function action that matches the operation command word based on the target device's communication address.

Images