Low-frequency signal detection method, control method and electronic device suitable for air puff recognition
By employing a dual decision mechanism of energy mutation detection and duration verification in low-frequency signal detection methods, combined with dynamic signal-to-noise ratio adjustment, the problems of false triggering and high hardware cost in the breath-blowing signal recognition of voice-controlled electronic devices are solved, achieving high accuracy and robust breath-blowing recognition.
Patent Information
- Application Number
- CN202510750458.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-06
- Publication Date
- 2025-12-09
- Estimated Expiration
- 2045-06-06
AI Technical Summary
Existing voice-controlled electronic devices suffer from poor sound source recognition, weak anti-interference capabilities, poor environmental adaptability, and high hardware costs when recognizing breath signals, resulting in high false trigger rates, response delays, and poor user experience.
A low-frequency signal detection method is adopted. By calculating the target frequency band energy difference and duration of the audio signal and combining it with dynamic signal-to-noise ratio adjustment, the accurate identification of the blowing signal is achieved. This includes a dual decision mechanism of energy change detection and duration verification, and the energy threshold is dynamically adjusted to adapt to environmental noise.
It significantly improves the accuracy of breath recognition and the robustness of the system, reduces the false trigger rate, enhances stability in complex acoustic environments, improves user experience, and reduces hardware costs.
Smart Images

Figure CN120375862B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of acoustic detection and electronic device control, and particularly to a low-frequency signal detection method and control method suitable for air blowing recognition and an electronic device. BACKGROUND
[0002] The current market acoustic control electronic devices mainly adopt the following technical solutions.
[0003] (1) Simple threshold detection method: the technical principle of this method is to compare the time domain signal amplitude collected by the microphone with a preset fixed threshold to realize triggering. The typical triggering implementation is to trigger the action when the signal amplitude exceeds the threshold (such as 1.5V). The main defects of this method include:
[0004] Poor sound source recognition ability: unable to distinguish different nature of sound sources such as air blowing, voice, collision, etc.
[0005] Weak anti-interference: sudden noise (such as the sound of falling objects) is easy to cause false triggering, and the actual measurement false triggering rate is as high as 18-25%;
[0006] Poor environmental adaptability: the fixed threshold cannot adapt to different environmental noise levels, and the missed triggering rate increases by more than 40% in noisy environments.
[0007] (2) Single frequency domain detection method: the technical principle of this method is to analyze the energy of a specific frequency band (such as 1kHz) using FFT. Its triggering method is to trigger when the energy of the target frequency band exceeds the threshold. The main defects of this method include:
[0008] Poor frequency adaptability;
[0009] Insufficient dynamic response: sensitive to sudden broadband noise (such as glass breaking sound), the actual measurement false triggering rate is 15-20%, and the response delay is generally more than 300ms.
[0010] (3) Hybrid sensor solution: the technical principle of this method is to combine the multi-modal detection of microphone and air pressure sensor (or infrared sensor), and trigger only when both acoustic signal and air pressure change conditions are met. The main defects of this method include:
[0011] Cost problem: increasing the air pressure sensor increases the BOM cost by 35-40%;
[0012] Reliability problem: the multi-sensor data fusion algorithm is complex, the failure rate is increased by 2-3 times, and the air pressure sensor is easily affected by environmental temperature and humidity, and the failure probability increases by 50% in humid environments;
[0013] User experience problem: detection distance is limited (usually <15cm), and sensitivity drops by 80% beyond the distance.
[0014] In summary, the low-frequency signal detection and control method in the prior art generally has at least the following deficiencies:
[0015] (1) Cannot distinguish specific frequencies, and is easily triggered by environmental noise (such as speech and music);
[0016] (2) Lack of anti-interference mechanism, poor stability in complex acoustic environment;
[0017] (3) Poor user experience: high response delay, unable to meet real-time interaction requirements, and lack of natural transition effect in state switching;
[0018] (4) High hardware cost: multi-sensor solution increases BOM cost by more than 35%.
[0019] The disclosure of the above background art is only used to assist in understanding the inventive concept and technical solutions of the present application, and does not necessarily belong to the prior art of the present application, nor does it necessarily provide technical teaching; in the absence of explicit evidence that the above content has been disclosed before the filing date of the present application, the above background art should not be used to evaluate the novelty and inventiveness of the present application. SUMMARY
[0020] The purpose of the present application is to provide a low-frequency signal detection method, control method and electronic device suitable for blow recognition, which can more accurately and reliably detect low-frequency signals based on blow.
[0021] To achieve the above purpose, the technical solutions adopted by the present application are as follows:
[0022] A low-frequency signal detection method suitable for blow recognition, for determining whether there is a low-frequency signal based on a blow signal in an audio signal, comprising the following steps:
[0023] Collecting an audio signal at a sampling frequency covering the target frequency band to obtain a to-be-detected audio signal;
[0024] Determining the signal energy E low (t) of the target frequency band corresponding to time t in the to-be-detected audio signal;
[0025] According to the signal energy E low (t) of the target frequency band, calculating the signal energy difference ΔE(t) in the time difference interval, ΔE(t) = E low (t) - E low (t-Δt), wherein Δt is a preset time difference interval;
[0026] Judging whether it satisfies: ΔE(t) < -Γ l and the duration of ΔE(t) < 0 meets the preset requirement, wherein Γ lΓ is an energy difference threshold l If the condition is not satisfied, it is determined that the audio signal to be detected does not include the low frequency signal based on the puff signal.
[0027] Further, any one of the above technical solutions or a combination of the above technical solutions further includes the following steps:
[0028] If the condition ΔE(t) < -Γ is satisfied l and the duration of ΔE(t) < 0 meets a preset requirement, it is determined that the audio signal to be detected includes the low frequency signal.
[0029] Further, any one of the above technical solutions or a combination of the above technical solutions further includes the following steps:
[0030] If the condition ΔE(t) < -Γ is satisfied l and the duration of ΔE(t) < 0 meets a preset requirement, a puff probability is calculated, if the puff probability is greater than a preset probability value, it is determined that the audio signal to be detected includes the low frequency signal, otherwise, it is determined that the audio signal to be detected does not include the low frequency signal.
[0031] Further, any one of the above technical solutions or a combination of the above technical solutions calculates the puff probability in the following manner:
[0032]
[0033] wherein P blow (t) represents the puff probability, σ(·) represents a probability conversion function, φ i (ΔE(t)) is a feature extraction function, and w i represents a weight coefficient.
[0034] Further, any one of the above technical solutions or a combination of the above technical solutions adopts a Sigmoid function as the probability conversion function, which is represented as follows: wherein e represents an index, and x is an independent variable.
[0035] and / or,
[0036] The feature extraction function adopts a Gabor base function, which is defined as: wherein x is an independent variable, μ i is a center position parameter, f i is a frequency adjustment parameter, ε i is a bandwidth adjustment parameter, cos(·) is a cosine function, π is a circular constant, i takes 1, 2, 3, and ψ i is a phase parameter.
[0037] Further, any one of the above technical solutions or a combination of the above technical solutions further comprises dynamically adjusting the weight coefficient w by the following manner i :
[0038] The real-time signal-to-noise ratio SNR(m) is determined by the following manner:
[0039]
[0040] wherein SNR(m) represents the real-time signal-to-noise ratio, ΔE(m) represents the normalized signal energy difference, m is a block index, σ N represents a noise standard deviation, and a is a proportional symbol;
[0041] According to the real-time signal-to-noise ratio SNR(m), the w is scaled by the following formula: i : w i ← w i · (1 + a · SNR(m)), wherein a represents a scaling coefficient.
[0042] Further, any one of the above technical solutions or a combination of the above technical solutions determines whether the duration of ΔE(t) < 0 meets the preset requirement by the following manner: if N S ≥ N th , it is determined that the duration of ΔE(t) < 0 meets the preset requirement, wherein N S represents a number of continuous frames meeting ΔE(t) < 0, and N th represents a preset frame number threshold.
[0043] Alternatively,
[0044] The duration of ΔE(t) < 0 is determined by the following manner:
[0045]
[0046] wherein k is a frame index, K is a preset continuous frame number threshold, z is a current starting frame, I(.) is an indicator function, which takes a value of 1 when the condition is met, and otherwise takes a value of 0, and β is a proportion coefficient, 0.5 ≤ β ≤ 1.
[0047] Further, any one of the above technical solutions or a combination of the above technical solutions calculates the signal energy E low (t) of a target frequency band in the audio signal to be detected at time t by the following manner:
[0048]
[0049] wherein X w(k, t) is a frequency domain representation of the audio signal to be detected after windowing processing, with dimensions of frequency points k x time frames t; SUM is the total number of FFT points, Corresponding to 1 / 4 bandwidth.
[0050] Further, any one of the technical solutions or a combination of the technical solutions described above further includes the following steps:
[0051] The sliding window is used to determine the signal energy difference △E(t) in a plurality of time difference intervals in the audio signal to be detected.
[0052] According to another aspect of the present application, an electronic device control method is provided, which determines whether there is a low-frequency signal based on a blowing signal in an audio signal based on the low-frequency signal detection method suitable for blowing recognition according to any one of the technical solutions or a combination of the technical solutions described above, and if so, controls the electronic device to switch from a first state to a second state, the second state being different from the first state.
[0053] Further, any one of the technical solutions or a combination of the technical solutions described above, the first state is an on state, and the second state is an off state.
[0054] According to another aspect of the present application, an electronic device is provided, which includes a controller configured to perform the electronic device control method according to any one of the technical solutions or a combination of the technical solutions described above, the electronic device being an electronic lamp and / or an electronic sound, the electronic lamp including an electronic candle and an electronic lamp stick.
[0055] Further, any one of the technical solutions or a combination of the technical solutions described above, the electronic device includes an electronic device body, an audio detection module, and a control module;
[0056] The audio detection module includes an audio acquisition module and a signal processing module, the audio acquisition module is configured to acquire an audio signal at a sampling frequency covering a target frequency band to obtain an audio signal to be detected, and the signal processing module is configured to determine whether there is a low-frequency signal based on a blowing signal in the audio signal to be detected and transmit the detection result to the control module.
[0057] The control module is configured to control the state of the electronic device body to change or not to change according to the detection result, including:
[0058] If the detection result is that there is a low-frequency signal based on a blowing signal in the audio signal to be detected, the state of the electronic device body is controlled to change, otherwise the state of the electronic device body is not controlled to change.
[0059] The technical solutions provided by the present application have the following beneficial effects:
[0060] a.The present application proposes a dual decision mechanism combining energy mutation detection and energy drop signal duration verification, which can significantly improve the accuracy of puff recognition and the robustness of the system. Energy mutation detection can quickly capture the energy drop caused by puff operation, while duration verification (the duration of signal energy drop meets the preset requirement) further confirms the persistence of the puff signal, ensuring the effectiveness and reliability of the puff signal. This cooperative detection method not only reduces the false trigger rate, but also enhances the stability of the system in complex acoustic environments;
[0061] b.The present application determines whether the duration of signal energy drop meets the preset requirement by whether the proportion of energy drop signal frame number meets the preset frame number threshold, which can avoid false triggering and false triggering caused by short-term interference, especially single-frame noise, ensuring the persistence of the puff signal and the reliability of the puff signal detection, further reducing the miss detection and false detection probability of the puff signal;
[0062] c.Based on the dual judgment mechanism, the present application further combines the judgment condition of puff probability, and adjusts the signal-to-noise ratio based on the environmental noise level to judge. According to the real-time signal-to-noise ratio (SNR), the energy threshold is dynamically adjusted, for example, when the environmental noise is significantly increased, the threshold is appropriately increased to reduce false triggering; on the contrary, in a quiet environment, the threshold can be appropriately reduced to improve the sensitivity, which can improve the robustness of the detection method and system in complex environments. BRIEF DESCRIPTION OF DRAWINGS
[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and those skilled in the art can obtain other drawings according to these drawings without creating any creative labor.
[0064] Figure 1 The flow chart of the first low-frequency signal detection method suitable for puff recognition provided by an exemplary embodiment of the present application;
[0065] Figure 2 The flow chart of the first method for determining whether the duration of signal energy being negative meets the preset requirement provided by an exemplary embodiment of the present application;
[0066] Figure 3 The flow chart of the second method for determining whether the duration of signal energy being negative meets the preset requirement provided by an exemplary embodiment of the present application;
[0067] Figure 4A flow chart of a second low-frequency signal detection method suitable for blow recognition provided for an exemplary embodiment of the present application is shown in Figure 2;
[0068] Figure 5 A module diagram of an electronic device controlled based on a low-frequency signal detection result provided for an exemplary embodiment of the present application is shown in Figure 3. DETAILED DESCRIPTION
[0069] In order to make the personnel in the technical field better understand the present application scheme, the technical scheme in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without making creative efforts should belong to the scope of protection of the present application.
[0070] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, device, product or apparatus including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or apparatuses.
[0071] In an embodiment of the present application, a low-frequency signal detection method suitable for blow recognition is provided, as shown in Figure 2, for determining whether a low-frequency signal based on a blow signal exists in an audio signal, comprising the following steps: Figure 1
[0072] An audio signal is collected at a sampling frequency covering the target frequency band to obtain a to-be-detected audio signal;
[0073] The signal energy E low (t) of the target frequency band corresponding to the time t in the to-be-detected audio signal is determined;
[0074] According to the signal energy E low (t) of the target frequency band, the signal energy difference ΔE(t) in the time difference interval is calculated, ΔE(t) = E low (t) - E low (t-Δt), wherein Δt is a preset time difference interval;
[0075] whether the following condition is satisfied: ΔE(t) < -Γ l and the duration of ΔE(t) < 0 satisfies a preset requirement, where Γ l is an energy difference threshold, Γ l > 0, and if not, it is determined that the audio signal to be detected does not include the low-frequency signal.
[0076] In an embodiment of the present application, the audio signal / data is collected at a sampling frequency F s ≥ 8 kHz, and the collected audio signal is windowed to reduce spectral leakage. The signal energy E low (t) of a target frequency band (for example, 0-1 kHz) in the audio signal to be detected at time t is calculated in the following manner:
[0077]
[0078] where X w (k, t) is a frequency domain representation of the audio signal to be detected after windowing, obtained by short-time Fourier transform (STFT), and has a dimension of frequency point k x time frame t; SUM is the total number of FFT points, for example corresponding to 1 / 4 bandwidth, i.e., 0-1 / 4 sampling rate range, and E low (t) has a unit of dB or linear amplitude square.
[0079] The calculation method of the signal energy E low (t) of the above target frequency band can focus on the core frequency band of the blowing signal (i.e., the target low-frequency band) and suppress high-frequency noise interference. In this embodiment, real-time signal energy monitoring is achieved by a sliding time window, which can adapt to the non-stationary signal characteristics.
[0080] The signal energy difference ΔE(t) in a plurality of time difference intervals in the audio signal to be detected is determined by using a sliding window, and the calculation formula of the signal energy difference ΔE(t) is as follows:
[0081] ΔE(t) = E low (t) - E low (t-Δt)
[0082] where Δt is a preset time difference interval, representing the time interval between two adjacent signal energy calculation times, and is used to calculate the signal energy change rate. Δ is a detection time window for determining the energy change, which is usually set according to the sampling frequency and the application scenario, for example, Δt can be 10ms-50ms.
[0083] In this embodiment, E low (t) represents the low-frequency band energy at time t (current time), i.e., the signal energy of the low-frequency band (such as 0-1 kHz) after filtering at time t, and Elow (t) reflects the intensity of the low-frequency band signal at the current time, which is the basis for calculating the energy change.
[0084] E low (t-Δt) represents the low-frequency band energy at a past time, specifically, it represents the energy of the low-frequency band signal after filtering at time t-Δt, E low (t-Δt) as a reference value, for comparison with the signal energy at the current time, to calculate the amount of change in signal energy.
[0085] The signal energy difference ΔE(t) is a signal energy mutation feature, which represents the amount of change in low-frequency band energy at time t, and is used to capture sudden changes in energy. By calculating the amount of change in energy, the energy drop caused by the blowing signal can be detected, so as to judge whether it is a valid blowing operation.
[0086] The calculation formula of the signal energy difference ΔE(t) is used to extract the signal energy mutation feature in the blowing signal. In the blowing operation, the airflow passing through the microphone will cause a sudden drop in low-frequency band energy. By calculating the energy difference ΔE(t) between the current time and a past time, this energy drop can be effectively captured, so as to judge whether it is a valid blowing signal.
[0087] For example, the blowing signal feature can be determined in the following way:
[0088] The energy quickly decays when blowing, and ΔE(t) is a significant negative value (such as -30dB);
[0089] The ΔE(t) fluctuation of environmental noise (such as speech) is small (such as ±5dB).
[0090] Based on the above-mentioned signal energy mutation feature in the blowing signal, the present application proposes ΔE(t) < -Γ l as one of the conditions for detecting the blowing signal. In addition, in order to improve the effectiveness of the blowing detection signal and reduce the false detection rate of the blowing signal, the present application further proposes that the duration of ΔE(t) < 0 meets the preset requirements as one of the conditions for detecting the blowing signal.
[0091] The present application proposes two methods to judge whether the duration of ΔE(t) < 0 meets the preset requirements, as shown in Figure 2 In the first judgment method:
[0092] If N S ≥ N th (Or Δt blow ≥ the preset time length), it is judged that the duration of ΔE(t) < 0 meets the preset requirements, wherein N S represents the number of consecutive frames that meet ΔE(t) < 0, Nth represents a preset frame number threshold for determining the frame number threshold of the energy change duration.
[0093] Based on the determination method, the condition for detecting an effective puff signal is as follows:
[0094] If ΔE(t) <-Γ l and N S ≥ N th , it is determined as an effective puff signal. Wherein, -Γl represents a preset negative energy threshold for determining the threshold of energy drop.
[0095] As Figure 3 shown, in the second determination method, different from the first determination method, whether the duration of ΔE(t) < 0 meets the preset requirement is determined by the following way:
[0096]
[0097] Wherein, k is the frame index, K is a preset continuous frame number threshold, represents the size of the detection window, i.e. the number of continuous frames that need to be verified, z is the current starting frame, and z+K represents the last frame of the detection window. For example, K = 10 represents the energy change in 10 frames, and the value of K is usually determined by experiment or experience, for example, the frame number corresponding to 100-300 ms is determined as K, which is converted into frame number by sampling rate, for example, when the sampling rate is 44.1 kHz and the frame length is 10 ms, K = 15 corresponds to a 150 ms detection window.
[0098] I(.) is an indicator function, which takes the value of 1 when the condition in () is true, otherwise 0.
[0099] The meaning of is to traverse each frame of the signal in the time window (such as the zth frame to the z+Kth frame), for checking the energy change ΔE(k) frame by frame whether it is negative. Puff signal usually leads to sudden drop of energy, and ΔE(k) < 0 is a key feature of mutation.
[0100] β is a proportion coefficient, 0.5 ≤ β ≤ 1. In this embodiment, take β = 0.5 as an example, satisfy Then it is determined that the duration of ΔE(t) < 0 meets the preset requirement.
[0101] Based on the determination method, the condition for detecting an effective puff signal is as follows:
[0102] If ΔE(t) <-Γ l and , it is determined as an effective puff signal.
[0103] Based on the condition of detecting the effective blowing signal, taking 0.5 as an example, the number of frames with energy reduction is required to be greater than or equal to K / 2 (that is, at least half of the frames need to meet the condition). Assuming K = 10, while meeting △E(t) < -Γ l , if only 3 frames of signals in the 10 frames of signals have energy reduction , it is determined that the energy reduction signal is noise; if 7 frames of signals in the 10 frames of signals have energy reduction , it is determined that the energy reduction signal is noise.
[0104] Through this method, the short-term interference can be avoided, especially the single-frame noise can be eliminated to cause the missed triggering and the false triggering, the continuity of the blowing signal and the reliability of the blowing signal detection are ensured, and the missed detection and false detection probability of the blowing signal is further reduced.
[0105] In an embodiment of the present application, a low-frequency signal detection method suitable for blowing recognition is provided, as shown in Figure 4 The low-frequency signal detection method suitable for blowing recognition provided by the embodiment further includes the following steps on the basis of the above-mentioned embodiment:
[0106] If the following conditions are met: △E(t) < -Γ l , and the duration of △E(t) < 0 meets the preset requirement, the blowing probability is calculated, if the blowing probability is greater than a preset probability value, it is determined that the audio signal to be detected includes the low-frequency signal, otherwise, it is determined that the audio signal to be detected does not include the low-frequency signal.
[0107] In the embodiment, the blowing probability is calculated in the following way:
[0108]
[0109] Wherein, P blow (t) represents the blowing probability, σ(·) represents the probability conversion function, φ i (△E(t)) is a feature extraction function, and w i represents a weight coefficient.
[0110] The preset signal energy threshold in the above-mentioned embodiment is usually determined according to a large amount of experimental data statistics, or determined according to experience, or determined based on standard learning samples by using AI algorithm, so as to ensure that the blowing signal and the noise can be effectively distinguished in a typical application scenario. For example, by analyzing the blowing data under different environmental noise conditions, an initial threshold is determined, so that the false triggering rate and the missed triggering rate are both at a low level, that is, an ideal level.
[0111] The embodiment is based on the above-mentioned embodiment and proposes a dynamic adjustment mechanism. In the detection process, the energy threshold is dynamically adjusted according to the real-time signal-to-noise ratio or other environmental parameters. For example, when it is detected that the environmental noise significantly increases, the threshold is appropriately increased to reduce false triggering; on the contrary, in a quiet environment, the threshold can be appropriately reduced to improve sensitivity.
[0112] Preferably, the probability conversion function adopts a Sigmoid function for mapping the feature value to the interval [0, 1] to output the blowing probability, and the probability conversion function σ(x) is represented as follows:
[0113]
[0114] wherein e represents an index, and x is an independent variable.
[0115] Preferably, the feature extraction function adopts a Gabor base function, which is defined as:
[0116]
[0117] wherein x is an independent variable, μ i is a center position parameter, f i is a frequency adjustment parameter, ε i is a bandwidth adjustment parameter, cos(·) is a cosine function, π is a circular constant, i takes 1, 2, 3, and ψ i represents a phase parameter, and its physical meaning is to control the initial phase of the cosine function. By adjusting ψ i , the phase characteristics of the signal can be changed, thereby affecting the time-frequency representation of the signal. It should be noted that the feature extraction function φ i (x) can be replaced by other time-frequency analysis base functions such as Morlet wavelet and Hermite function in addition to the Gabor base function.
[0118] In the embodiment, the physical meaning, typical value and determination method of the related parameters involved in the dynamic adjustment mechanism are shown in Table 1.
[0119] Table 1 Related parameters of dynamic adjustment mechanism
[0120] Parameter Physical meaning Typical value Determination method μ i ]]> Center position parameter 0.5,1.0,1.5 Matching blowing signal characteristic peak e i ]] Bandwidth adjustment parameter 0.2~0.5 Optimized by cross-validation f i ]]> Frequency adjustment parameter 10Hz, 20Hz Set as the harmonic of the blowing signal fundamental frequency
[0121] Preferably, the weight coefficient w i is dynamically adjusted in the following manner to adapt to different detection environments.
[0122] The real-time signal-to-noise ratio SNR(m) is determined according to the following formula: wherein SNR(m) represents the real-time signal-to-noise ratio, ΔE(m) represents the normalized signal energy difference, m is a block index, σ NThe noise standard deviation is denoted by σ, which is obtained by the ambient noise estimation module. The greater the absolute value of ΔE(m), the more significant the signal energy change, and the higher the signal-to-noise ratio SNR.
[0123] The calculation formula of ΔE(m) is:
[0124]
[0125] wherein σ E denotes the sliding mean, μ E denotes the sliding standard deviation. The differential feature extraction can standardize the energy change, improve the robustness of the feature, and make it not affected by the absolute energy level, so as to more effectively capture the energy change. The mean and standard deviation are calculated through the sliding window, the energy feature is standardized in real time, and the anti-interference ability of the system is enhanced.
[0126] E l The calculation formula of ΔE(m) is as follows:
[0127]
[0128] wherein L is the block length (such as 256 samples), which is standardized by the sliding mean and standard deviation to enhance the robustness of the detection system. ΔE(m) is the standardized signal energy difference based on the discrete signal block, which is used for standardized feature extraction and improves the anti-interference ability through statistical normalization. The core role of ΔE(m) is to normalize the energy change through the sliding mean and standard deviation, which is suitable for the input of the probability model (such as P blow (t) calculation) and the robustness decision in complex environments (such as dynamic adjustment of threshold), which can eliminate the influence of the absolute energy level and adapt to the dynamic environment (such as the change of signal-to-noise ratio). It is not suitable for real-time mutation detection because its calculation depends on block statistics (such as 256 samples / block), which will introduce delay.
[0129] According to the real-time signal-to-noise ratio SNR(m), w i is scaled by a proportion: i ← w i · (1 + α · SNR(m)), wherein 1 + α · SNR(m) is an adjustment factor for dynamically scaling the weight coefficient according to the signal-to-noise ratio; 1 ensures that the weight coefficient at least maintains the original value and will not be reduced because the signal-to-noise ratio is negative (although the signal-to-noise ratio is usually non-negative), α represents a proportion coefficient, which determines the degree of influence of the signal-to-noise ratio on the weight coefficient, and α can be adjusted according to the requirements of system design, and SNR represents the signal-to-noise ratio, i.e. the ratio of signal power to noise power. In practical applications, w i can be fitted by least squares method through the collection of multiple groups of blowing data in a quiet environment. In addition, the weight coefficient w i can also be dynamically optimized through online learning algorithm (such as LMS adaptive filtering).
[0130] In the embodiment, the weight coefficient w is dynamically adjusted i The parameter meanings of the related formulas are shown in Table 2.
[0131] Table 2 Parameter meanings in the related formulas of dynamically adjusting the weight coefficient
[0132] Parameter Definition Description E l (m)]]> Energy of the mth block Indicates the sum of signal energy from time point mL to (m+1)L-1 m Block index Used to identify the currently processed signal block L Block length The number of samples contained in each signal block n Sample index Sample position in the signal sequence x(n) Signal sample value Amplitude of the input signal at time point n
[0133] Since the SNR signal-to-noise ratio needs to rely on the block statistical results, the formula based on ΔE(m) can assist in explaining the source of the signal-to-noise ratio data in the dynamic adjustment of the weight coefficient. The probability model is directly input as the real-time feature ΔE(t), while ΔE(m) is used for system-level parameter optimization (such as weight adjustment) and does not participate in probability calculation, that is, the signal-to-noise ratio SNR may rely on the block statistical results of ΔE(m), thereby indirectly affecting the output of P blow (t).
[0134] As described above, the low-frequency signal detection method suitable for blow recognition proposed by the present application proposes a double decision mechanism combining energy mutation detection and duration verification, which can significantly improve the accuracy of blow recognition and the robustness of the system. Energy mutation detection can quickly capture the energy drop caused by blow operation, and duration verification further confirms the persistence of the blow signal, ensuring the effectiveness and reliability of the blow signal. This cooperative working mode not only reduces the false trigger rate, but also enhances the stability of the system in complex acoustic environments.
[0135] For example, in a specific embodiment, the joint decision condition can be set as follows:
[0136]
[0137] Wherein:
[0138] Trigger off (t) represents that the blow signal is detected;
[0139] is an indicator function, used to judge whether the condition is met or not;
[0140] Δt blow is the duration of the blow signal, used to ensure the persistence of the blow signal;
[0141] [T min ,T max ] is the duration range of the blow signal, determined by experimental statistics, and the typical value is 100-300 ms.
[0142] Through the cooperative working mechanism, the application can verify whether the duration of the energy mutation meets the characteristics of the blowing signal while capturing the energy mutation, so as to effectively distinguish the blowing signal from environmental noise or other interference signals. The double verification mechanism significantly improves the accuracy of blowing recognition and the robustness of the system.
[0143] In an embodiment of the application, an electronic device control method is provided, which determines whether a low-frequency signal based on a blowing signal exists in an audio signal based on the low-frequency signal detection method suitable for blowing signal recognition according to any one of the above, and if so, controls the electronic device to switch from a first state to a second state, the second state being different from the first state.
[0144] In actual application, the electronic device can be electronic products such as electronic lamps such as electronic candles and electronic sound, etc. The first state and the second state can be one of the states of off and the other state of on, or two states of different brightness or display patterns or light colors.
[0145] In an embodiment of the application, an electronic device is provided, which includes a controller configured to execute the electronic device control method according to the above embodiments. The electronic device is an electronic lamp and / or an electronic sound, and the electronic lamp includes an electronic candle and an electronic lamp stick.
[0146] In the embodiment, as shown in the figure, the electronic device includes an electronic device body, an audio detection module, a control module and a visual feedback module. Figure 5
[0147] The audio detection module includes an audio acquisition module and a signal processing module. The audio acquisition module is configured to acquire an audio signal at a sampling frequency covering a target frequency band to obtain an audio signal to be detected. The signal processing module is configured to determine whether the audio signal to be detected has a low-frequency signal based on a blowing signal, and transmit a detection result to the control module.
[0148] The control module is configured to control the state of the electronic device body to change or not to change according to the detection result, including:
[0149] If the detection result is that the audio signal to be detected has a low-frequency signal based on a blowing signal, the state of the electronic device body is controlled to change, otherwise the state of the electronic device body is not controlled to change.
[0150] The visual feedback module includes a display screen. If the detection result is that the audio signal to be detected has a low-frequency signal based on a blowing signal, the display screen is controlled to display a pattern according to a preset rule.
[0151] In one specific embodiment of the present application, the hardware configuration of the electronic device is shown in Table 3.
[0152] Table 3 Hardware configuration of the electronic candle
[0153] Module Parameter Embodiment configuration Microphone Omnidirectional microphone Frequency response: 20Hz-20kHz ADC ≥12-bit resolution Sampling rate: 44.1kHz Master chip Microcontroller supporting FFT calculation Operating frequency ≥100MHz
[0154] The performance test data of the electronic device is shown in Table 4.
[0155] Table 4 Performance test data of the electronic candle
[0156] Indicator Test condition Result Detection accuracy Target frequency sine wave, 60dB environmental noise >99% False trigger rate Human voice, music interference (<10 kHz) <0.5% Response delay From sound wave input to trigger execution ≤200ms
[0157] The parameter optimization range of the electronic device is shown in Table 5.
[0158] Table 5 Parameter optimization range of the electronic device
[0159] Parameter Meaning Preferred range F s ]]> Sampling frequency 8k-48kHz N FFT point number 256-1024 Δf Frequency tolerance 50-500Hz M Continuous detection times 5-20 times [CAT c ]]> Cooling time 500-2000ms
[0160] The present application relates to low-frequency signal detection and intelligent electronic device control technology, through double decision logic, realizing high robustness non-contact interaction control, especially suitable for high-precision trigger demand of electronic candle, intelligent lighting and other scenes.
[0161] In one specific embodiment of the present application, the electronic device is an electronic candle with an LED screen. When blowing the electronic candle, the microphone detects the low-frequency signal generated by blowing, and then controls the electronic candle to be extinguished. The specific extinguishing method can be gradually extinguished from top to bottom in the LED screen by combining a non-linear brightness simple method with a row scanning mask, or gradually darkened to extinguish in a specific pattern, or playing music while extinguishing the electronic candle.
[0162] It should be noted that the above electronic device control method, electronic device embodiment and low-frequency signal detection method suitable for blowing recognition embodiment are based on the same inventive concept. The entire content of the low-frequency signal detection method suitable for blowing recognition embodiment is incorporated by reference into the electronic device control method and the electronic device embodiment.
[0163] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0164] The above description is only a specific embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A low-frequency signal detection method suitable for breath-blowing recognition, characterized in that, To determine whether a low-frequency signal based on a blowing signal is present in an audio signal, the following steps are included: Audio signals are acquired at sampling frequencies covering the target frequency band to obtain the audio signal to be detected; Determine the signal energy E of the target frequency band at time t in the audio signal to be detected. low (t); According to the signal energy E of the target frequency band low ΔE(t) is used to calculate the signal energy difference within the time difference interval. Where △t is the preset time difference interval; Determine if the following conditions are met: and The duration meets the preset requirements, where, The energy difference threshold, If the condition is not met, then it is determined that the audio signal to be detected does not include the low-frequency signal; If the following conditions are met: and If the duration of the signal meets the preset requirement, the blowing probability is calculated. If the blowing probability is greater than the preset probability value, it is determined that the audio signal to be detected includes the low-frequency signal; otherwise, it is determined that the audio signal to be detected does not include the low-frequency signal. The blowing probability is calculated in the following way: ; in, Indicates the probability of blowing air. This represents the probability transformation function. w is the feature extraction function. i Indicates the weighting coefficient; The weighting coefficient w is dynamically adjusted in the following manner. i The real-time signal-to-noise ratio (SNR) (m) is determined as follows: ; Where SNR(m) represents the real-time signal-to-noise ratio. Represents the normalized signal energy difference. For block indexes, Indicates the standard deviation of noise. It is a proportional sign; Scaled proportionally based on real-time signal-to-noise ratio (SNR) (m) : ,in, This represents the proportionality coefficient.
2. The low-frequency signal detection method for breath recognition according to claim 1, characterized in that, It also includes the following steps: If the following conditions are met: and If the duration of the signal meets the preset requirements, then the audio signal to be detected is determined to include the low-frequency signal.
3. The low-frequency signal detection method for breath recognition according to claim 1, characterized in that, The probability transformation function is the Sigmoid function, which is expressed as follows: , where e represents the exponent and x is the independent variable; And / or, The feature extraction function uses Gabor basis functions and is defined as follows: ,in, As the independent variable, For the center position parameter, For frequency adjustment parameters, For bandwidth adjustment parameters, Let i be the cosine function, π be the mathematical constant pi, and i take the values 1, 2, or 3. This is the phase parameter.
4. The low-frequency signal detection method for breath recognition according to claim 1, characterized in that, Judged by the following methods Does the duration meet the preset requirements? Then judge The duration meets the preset requirements, where, Indicates satisfaction The number of consecutive frames, This indicates the preset frame rate threshold; or, Judged by the following methods Does the duration meet the preset requirements? ; Where k is the frame index, K is the preset continuous frame count threshold, and z is the current starting frame. The period (.) is an indicator function that takes the value 1 if the condition is true and 0 otherwise. This is the percentage coefficient. .
5. The low-frequency signal detection method for breath recognition according to claim 1, characterized in that, The signal energy E of the target frequency band in the audio signal to be detected at time t is calculated as follows: low (t): ; in, This is the frequency domain representation of the audio signal to be detected after windowing, with its dimension being frequency points. ×Timeframe ; The total number of points in the FFT. Corresponding to 1 / 4 bandwidth.
6. The low-frequency signal detection method for breath recognition according to claim 1, characterized in that, It also includes the following steps: A sliding window is used to determine the signal energy difference ΔE(t) within multiple time difference intervals in the audio signal to be detected.
7. A method for controlling an electronic device, characterized in that, Based on the low-frequency signal detection method for breath recognition as described in any one of claims 1 to 6, determine whether there is a low-frequency signal based on a breath signal in the audio signal; if so, control the electronic device to switch from a first state to a second state, the second state being different from the first state.
8. An electronic device, characterized in that, The electronic device includes a controller configured to perform the electronic device control method as claimed in claim 7, wherein the electronic device is an electronic lamp and / or an electronic sound, and the electronic lamp includes an electronic candle and an electronic lamp stick.
Citation Information
Patent Citations
User input
WO2011082535A1