Multi-scene correction method for measuring room temperature based on dual-channel microphone intelligent equipment

Abstract

The invention discloses a multi-scene correction method for measuring room temperature based on dual-channel microphone intelligent equipment. The multi-scene correction method comprises the following steps: S1, calling an equipment loudspeaker to emit a sound signal; s2, calling a sound signal and preprocessing an original dual-channel sound signal; s3, determining arrival sounds of different paths, and screening interference signals; s4, calculating the signal receiving time delay of the dual-channel microphone; s5, estimating the sound velocity based on a linear acoustic hypothesis by using the shortest sound path distance difference and the time delay difference of the sound waves, and obtaining the average temperature of the sound path; and S6, for different use environments and equipment models, initialization is carried out before first use. According to the method, the calculation efficiency when the algorithm is deployed and operated on the intelligent equipment can be remarkably improved, the accuracy and stability of a temperature measurement result can be improved for a complex temperature measurement scene and external interference, and the algorithm can obtain a good temperature measurement result on various types of equipment in multiple scenes.

Description

Technical Field

[0001] The invention relates to a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device, and belongs to the technical field of signal processing. Background Art

[0002] The measurement of room temperature has important application value in the fields of space comfort assessment and clinical monitoring analysis. If some portable smart devices can be used for real-time and on-site temperature measurement, it will not only bring great convenience to room temperature measurement, but also reduce the measurement cost. However, the current means of measuring room temperature based on equipment is mainly based on infrared temperature measurement, which requires the installation of additional hardware and is not ideal in terms of accuracy.

[0003] From an acoustic perspective, the device speaker can be used to emit sound waves, and the sound speed can be calculated by measuring the delay of the dual-channel microphone receiving the signal and combining the length of each sound propagation path. Then, the room temperature can be estimated based on the mapping relationship between sound speed and temperature. However, there is usually noisy background noise (human voice, wind sound, etc.) in the temperature measurement scenario, and the multipath propagation of sound waves will interfere with the delay estimation of the dual-channel signal. The common delay estimation algorithm has a large amount of calculation and data. The temperature measurement efficiency may be low when actually deployed on the device, and the influence of noise and multipath effects will bring great challenges to the robustness of the algorithm. Therefore, designing a lightweight and robust data processing method is of great significance for the application of ultrasonic measurement of room temperature of smart devices. Summary of the invention

[0004] The purpose of the present invention is to propose a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device. This method can effectively reduce the calculation amount and data volume of the general sound speed-room temperature measurement algorithm based on time delay estimation, and improve the accuracy and robustness of the temperature measurement results in a noisy environment.

[0005] To achieve the above object, the technical solution adopted by the present invention is: a multi-scenario correction method for measuring room temperature based on a dual-channel microphone smart device, which comprises the following steps:

[0006] S1: Call the device speaker to emit a sound signal;

[0007] S2: Call the dual-channel microphone of the device to collect sound signals and pre-process the original dual-channel sound signals;

[0008] S3: Use the excitation signal and the dual-channel sound signal to perform cross-correlation calculation to determine the sound arriving from different paths and filter out interference signals;

[0009] S4: Calculate the time delay of the dual-channel microphone receiving signal by calculating the time delay between the excitation signal and the shortest direct sound of the dual-channel sound signal screened out;

[0010] S5: Estimate the sound speed based on the shortest acoustic path difference and time delay difference of sound waves, and obtain the average temperature of the sound path according to the sound speed;

[0011] S6: For different usage environments and device models, perform a temperature measurement once before the first use to complete initialization. Specifically, first perform a temperature measurement through steps S1 - S5, then reverse - infer the size parameters of the device and update the algorithm parameters based on the temperature reference values measured by other existing temperature measurement methods, and then use the updated parameters for subsequent temperature measurements.

[0012] Further, in step S1, the transmitted signal can adopt a segment of linear frequency - modulated signal, which can reduce the propagation sound attenuation without increasing the signal length and ensure the spatial resolution. The center frequency of the signal is preferably above 18 kHz or in a higher ultrasonic frequency band to avoid environmental noise interference, and at the same time, set a certain bandwidth to make up for the frequency response differences of the device. Considering the possible system delays in starting the transmitting and receiving threads of the device and the device shaking caused by the user during operation, set a certain blank paragraph at the beginning of the excitation signal, set the acquisition length to be the same as the length of the excitation signal, and ensure that the acquired signal includes the complete excited part and is relatively stable. When acquiring, to avoid audio distortion and ensure the calculation accuracy, the audio format is 16 - bit depth, and set the buffer size:

[0013]

[0014] In the above formula, f s is the sampling frequency of the device microphone. Setting the buffer size through the above formula can ensure calculation while sampling and control the amount of calculated data, so as to simultaneously perform the signal transmitting, receiving, and calculating threads, and maximize the calculation efficiency through parallel calculation.

[0015] Further, considering the influence of the multi - path propagation effect of sound waves on the time - delay calculation, it is necessary to strictly limit the synchronous end of the transmitting thread and the receiving thread. Missing a small part of the end signal has no significant impact on the time - delay estimation. Therefore, step S2 pre - processes the dual - channel signal according to the following steps:

[0016] S2.1: Let the length of the reserved blank paragraph in S1 be t0, and the total length of the signal be t a , perform peak search on the first t0 segments of the received signal, and use the peak s m0 to normalize the signal;

[0017] S2.2: Calculate the ratio h of the maximum peak s m0 to the second - largest peak s m1 . According to the experiment, set a threshold h0. If h < h0, it is considered that the background noise is relatively stable, and from the received signal The signal is intercepted from the moment h ≥ h0; if h ≥ h0, it is considered that there is strong interference in the environment, and the maximum peak value s m0 Intercept the signal after the moment;

[0018] S2.3: Use FIR bandpass filter to reduce noise on the intercepted signal, and the filter passband is determined by the frequency band of the excitation signal;

[0019] S2.4: Let the interception starting point in S2.2 be t c , with [t c ,t0] part of the signal is used as a reference, and spectral subtraction is used to a ] segment signal noise reduction;

[0020] S2.5: Since the interception operation of the dual-channel signal in S2.2 may be different, the length of the corresponding signal may be different, and the corresponding stimulus response time is also different, which will cause the delay between the received signal and the stimulus signal to deviate, so the signal must be compensated. According to the previous interception length, add the corresponding long 0 at the beginning of the signal so that the length of the dual-channel signal is still t a ;

[0021] S2.6: Based on the length l of the signal sequence in S2.5, calculate the smallest positive integer n such that 2 n ≥l, fill in 0 from the end of the signal to make the signal length 2 n , and then up-sampled, and after up-sampling, passed through a bandpass filter to improve the calculation efficiency and avoid spectrum leakage of the signal.

[0022] Furthermore, in step S2.1, the blank segment length t0 depends on the performance of the device and the temperature measurement environment. Generally speaking, to ensure the universality of the algorithm,

[0023] Furthermore, in the steps S3-S5, assuming that the difference in acoustic path length is d, and the time delays between the dual-channel signal and the excitation signal are τ1 and τ2 respectively, the ambient sound speed is:

[0024]

[0025] Furthermore, the average ambient temperature of the sound path is obtained by the following formula:

[0026]

[0027] Among them, T0 = 273.15K, c0 = 331.5m / s is the sound speed corresponding to T0, and the ambient temperature can be obtained by substituting the sound speed in formula (1) into formula (2).

[0028] Furthermore, in step S3, the time delay of the sound arriving from different paths is determined by calculating the cross-correlation peak between the excitation signal and the dual-channel signal, and the sound with the shortest time delay is selected as the sound signal of the known path d (it is assumed that the length of the shortest propagation path is the distance between the two microphones of the device, and this assumption can be verified through experiments).

[0029] Furthermore, in step S5, due to the limitations of the sound velocity-room temperature measurement algorithm based on time delay estimation, the results of the algorithm for room temperature estimation are obviously different for different usage scenarios and device models. For example, if the device is in a state of intense motion during signal transmission and reception, the sound velocity near the device will be affected by changes in air density and produce large fluctuations. The resulting signal distortion and changes in the environment itself will directly lead to abnormal temperature measurement results. Therefore, when measuring room temperature, it is necessary to ensure that the entire device is in a relatively standardized environment, such as placed horizontally on a desktop, etc.; in addition, the estimation of the sound velocity is related to the difference in the length of the acoustic path from the speaker to the two microphones. For different models of equipment, this parameter needs to be determined before measurement.

[0030] In summary, for the room temperature measurement algorithm that needs to be deployed on the device, initialization before the first use is necessary, which comes from both the theoretical basis of the algorithm and the considerations of the algorithm application. In addition, some signal filtering or smoothing methods can be adopted to improve the temperature measurement accuracy and algorithm robustness during the temperature measurement process. The initialization in S6 is essentially to use the algorithm proposed by the present invention to measure the temperature first, and then infer the environmental parameters used in the algorithm based on the algorithm results and other external temperature measurement reference values. The specific methods and processing strategies are as follows:

[0031] Consider temperature measurement in three scenarios: horizontal desktop, one-handed vertical, and two-handed horizontal:

[0032] A1: For a horizontal desktop, the temperature measurement scenario is relatively simple and stable, and the sound propagation path is relatively clear. Only one temperature measurement is required, and the temperature reference value T is input from the outside (which can be determined by other temperature measurement methods). R And the reference sound speed c is inferred from formula (2) R , and then calculate the corrected mobile phone size parameters based on the delay τ2-τ1 calculated during the temperature measurement process. The subsequent temperature measurement process uses d ′ Just replace d;

[0033] A2: For one-handed vertical holding, use the same method as A1 for initialization. In addition, the shaking of the user's arm needs to be considered. The Kalman filter algorithm can be used to smooth the data and improve the accuracy of temperature estimation. Suppose the temperature estimate Z(k) at time k is:

[0034] Z(k)=T(k)+v(k)

[0035] The evolution equation of the real temperature T(k) over time is:

[0036] T(k)=T(k-1)+w(k-1)

[0037] Where v and w are the measurement noise and system noise, which are set to be Gaussian white noise with mean 0 and variance Q and R. Assuming the measurement error covariance is P, then considering that the room temperature is close to constant when the temperature is measured continuously (the temperature measurement interval is less than 3s), we have:

[0038] T(k|k-1)=T(k-1|k-1)

[0039] P(k|k-1)=P(k-1|k-1)+Q

[0040] Wherein, T(k|k-1) is the predicted temperature at time k at time k-1, and P(k|k-1) is the prediction error covariance.

[0041] Calculate the Kalman gain K(k):

[0042]

[0043] Then there is an updated temperature estimate:

[0044] T(k|k)=T(k|k-1)+K(k)[Z(k)-T(k|k-1)]

[0045] Update the error covariance for subsequent measurements:

[0046] P(k|k)=[1-K(k)]P(k|k-1)

[0047] Kalman filtering can effectively reduce the uncertainty in temperature measurement caused by noise and handle random external factors to a certain extent. It does not require tracing historical data and can effectively reduce dependence on computing power.

[0048] A3: For holding both hands horizontally, use the same method as A1 for initialization. Since the human hand has significantly blocked the shortest sound path in the horizontal state, and the thermal radiation of the human hand will increase the temperature of the local area near the sound path (experiments have verified that this effect increases the measured value by about 0.5 to 1°C, which is quite considerable), in addition to using the Kalman filter algorithm described in A2, a fine adjustment can be made to the measurement result:

[0049]

[0050] in is the corrected temperature, This is the measurement result before correction.

[0051] The beneficial effects of the present invention are as follows: the present invention proposes a signal preprocessing method and a processing strategy for real-time correction of the algorithm in multiple scenarios and multiple models, which can significantly improve the computing efficiency of the algorithm when it is deployed and run on smart devices, and can improve the accuracy and stability of the temperature measurement results for complex temperature measurement scenarios and external interference, so that the algorithm can obtain good temperature measurement results on multiple models of equipment in multiple scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a schematic diagram of the ultrasonic propagation path of the device;

[0053] Figure 2 This is the flow chart of the temperature measurement algorithm;

[0054] Figure 3 Flowchart for temperature measurement and initialization (calibration) deployed on the device;

[0055] Figure 4 This is the flow chart of Kalman filter algorithm;

[0056] Figure 5 is the temperature estimation result after Kalman filtering DETAILED DESCRIPTION

[0057] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0058] Embodiment 1

[0059] This embodiment proposes a multi-scenario correction method for measuring room temperature based on a dual-channel microphone smartphone, comprising the following steps:

[0060] S1: Call the device speaker to emit a sound signal;

[0061] S2: Call the dual-channel microphone of the device to collect sound signals, and perform pre-processing such as cropping and noise reduction on the original dual-channel sound signals;

[0062] S3: Use the excitation signal and the dual-channel signal to perform cross-correlation calculation to determine the arrival of sounds from different paths and filter out interference signals;

[0063] S4: Calculate the time delay of the dual-channel microphone receiving signal by calculating the time delay between the excitation signal and the shortest direct sound of the dual-channel signal screened out;

[0064] S5: using the difference in the shortest sound path and the time delay of the sound waves, the speed of sound is estimated based on the linear acoustic assumption;

[0065] S6: For different usage environments and device models, in order to ensure algorithm performance, it is necessary to perform a temperature measurement through steps S1-S5 before the first use after the algorithm is deployed (i.e., steps S1-S5), and then use the temperature reference value measured by other methods (such as a thermocouple thermometer) to reversely infer the phone size parameters and update the algorithm parameters. In addition, corresponding processing strategies are adopted for different scenarios.

[0066] by Figure 1 , Figure 2 , Figure 3 and Figure 4 Taking the ultrasonic temperature measurement task of the equipment shown as an example, the technical points contained in the method of the present invention are explained in detail as follows:

[0067] 1. Sound velocity measurement based on time delay estimation method:

[0068] like Figure 1 As shown in the figure, for a general smartphone with a dual-channel microphone, the excitation signal is emitted from the speaker at the bottom right of the device and has multiple propagation paths to the bottom left and top right of the device. The two shortest acoustic paths with clear lengths d are used. 20 and d 10 To calculate the time delay, the acoustic path length difference is:

[0069] d=d 20 -d 10

[0070] The speed of sound can be calculated based on the time delay difference τ2-τ1:

[0071]

[0072] The average temperature of the acoustic path is obtained by the following formula:

[0073]

[0074] Among them, T0=273.15K, c0=331.5m / s is the sound speed corresponding to T0.

[0075] 2. Measurement and initialization using a smartphone:

[0076] like Figure 2 -a, after the measurement process starts, the signal transmission and reception threads are started simultaneously. If the calculation accuracy is expected to be further improved, the processing thread can be started some time before the end of the reception. Part of the calculation in S2 can be performed synchronously with the reception. The initialization steps are as follows: Figure 2 -b, the basic idea is to update the relevant parameters in the algorithm (mainly the mobile phone size parameter d) according to the delay estimation in the current scenario after a temperature measurement.

[0077] Initialization essentially means first using the algorithm proposed by the present invention to measure temperature, and then inferring the environmental parameters used in the algorithm based on the algorithm results and other external temperature measurement reference values. The specific method and processing strategy are as follows:

[0078] Consider temperature measurement in three scenarios: horizontal desktop, one-handed vertical, and two-handed horizontal:

[0079] A1: For a horizontal desktop, the temperature measurement scenario is relatively simple and stable, and the sound propagation path is relatively clear. Only one temperature measurement is required, and the temperature reference value T is input from the outside (which can be determined by other temperature measurement methods). R And the reference sound speed c is inferred from formula (2) R , and then calculate the corrected phone size parameters based on the delay τ2-τ1 calculated during the temperature measurement process:

[0080] d ′ =c R (τ2-τ1)

[0081] The subsequent temperature measurement process uses d ′ Just replace d.

[0082] A2: For one-handed vertical holding, use the same method as A1 for initialization. In addition, the shaking of the user's arm needs to be considered. The Kalman filter algorithm can be used to smooth the data and improve the accuracy of temperature estimation. Measure the temperature continuously, and assume that the temperature estimate Z(k) at time k is:

[0083] Z(k)=T(k)+v(k)

[0084] The evolution equation of the real temperature T(k) over time is:

[0085] T(k)=T(k-1)+w(k-1)

[0086] Where v and w are the measurement noise and system noise, which are set to be Gaussian white noise with mean 0 and variance Q and R. Assuming the measurement error covariance is P, then considering that the room temperature is close to constant when the temperature is measured continuously (the temperature measurement interval is less than 3s), we have:

[0087] T(k|k-1)=T(k-1|k-1)

[0088] P(k|k-1)=P(k-1|k-1)+Q

[0089] Wherein, T(k|k-1) is the predicted temperature at time k at time k-1, and P(k|k-1) is the prediction error covariance.

[0090] Calculate the Kalman gain K(k):

[0091]

[0092] Then there is an updated temperature estimate:

[0093] T(k|k)=T(k|k-1)+K(k)[Z(k)-T(k|k-1)]

[0094] Update the error covariance for subsequent measurements:

[0095] P(k|k)=[1-K(k)]P(k|k-1).

[0096] A3: For holding both hands horizontally, use the same method as A1 for initialization. Since the human hand has significantly blocked the shortest sound path in the horizontal state, and the thermal radiation of the human hand will increase the temperature of the local area near the sound path (experiments have verified that this effect increases the measured value by about 0.5 to 1°C, which is quite considerable), in addition to using the Kalman filter algorithm described in A2, a fine adjustment can be made to the measurement result:

[0097]

[0098] in is the corrected temperature, This is the measurement result before correction.

[0099] 3. Temperature estimation based on Kalman filter algorithm:

[0100] like Figure 4 As shown in the figure, Kalman filtering is an algorithm that makes the best estimate of the system state based on the prior model and observation data of the system. Its core idea is to use the observation data and model to predict the state of the system at the next moment when there is a certain understanding of the system, that is, to master the dynamic model of the system, and then correct the new observation data at the next moment by the difference between the new observation data and the predicted value at the previous moment, so as to obtain the best estimate of the state of the system at this moment.

[0101] For temperature prediction, continuous measurements can be performed when the room temperature is approximately constant for a short period of time. The Kalman gain K(k) is used to selectively absorb the current measurement result according to the corresponding state equation. If the current result is significantly different from the previous one, the observation and prediction are weighted averaged according to the environment and measurement noise. The larger the K(k), the more trust is given to the new observation. The impact of occasional interference on temperature measurement is minimized. The initial state estimate and error can be obtained by experiment.

[0102] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the protection scope of the present invention in any form, and all technical solutions obtained by equivalent replacement and the like fall within the protection scope of the present invention. The parts not involved in the present invention are the same as the prior art or can be implemented by the prior art.

Claims

1. A multi-scenario correction method for measuring room temperature based on a dual-channel microphone smart device, characterized in that: The steps include: S1: Call the device speaker to emit a sound signal; S2: Call the dual-channel microphone of the device to collect sound signals and pre-process the original dual-channel sound signals; S3: Use the excitation signal and the dual-channel sound signal to perform cross-correlation calculation to determine the sound arriving from different paths and filter out interference signals; S4: Calculate the time delay of the dual-channel microphone receiving signal by calculating the time delay between the excitation signal and the shortest direct sound of the dual-channel sound signal screened out; S5: Using the difference in the shortest sound path and the time delay of the sound wave, the sound velocity is estimated based on the linear acoustic assumption, and the average temperature of the sound path is obtained according to the sound velocity; S6: For different usage environments and equipment models, a temperature measurement is performed before the first use to complete initialization. Specifically, a temperature measurement is performed once through steps S1-S5, and then the size parameters of the equipment are inferred and the algorithm parameters are updated based on the temperature reference value measured by other existing temperature measurement methods, and then the updated parameters are used for subsequent temperature measurements.

2. According to claim 1, a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device is characterized in that: In step S1, the transmitted sound signal adopts a linear frequency modulation signal, the center frequency of the signal is in the ultrasonic frequency band above 18kHz, and a certain bandwidth is set to compensate for the frequency response difference of the equipment, and a blank section of a certain length is designed at the beginning.

3. According to claim 2, a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device is characterized in that: In step S2, the dual-channel signal is preprocessed according to the following steps: S2.1: Assume that the length of the blank segment reserved in S1 is t0, and the total length of the signal is t a , perform peak search on the first t0 segment of the received signal, using the peak value s m0 Normalize the signal; S2.2: Calculate the maximum peak value s m0 and the secondary peak value s m1 The ratio h is calculated. According to the experimental setup, a threshold h0 is set. If h < h0, it is considered that the background noise is relatively stable, and the signal is intercepted from the time of the received signal; if h ≥ h0, it is considered that there is strong interference in the environment, and the signal is intercepted after the time when the maximum peak value s m0 appears; S2.3: Use FIR bandpass filter to reduce noise on the intercepted signal, and the filter passband is determined by the frequency band of the excitation signal; S2.4: Let the interception starting point in S2.2 be t c , with [t c ,t0] part of the signal is used as a reference, and spectral subtraction is used to a ] segment signal noise reduction; S2.5: Add the corresponding length of 0 at the beginning of the signal according to the previous interception length, so that the length of the dual-channel signal is still t a ; S2.6: Based on the length l of the signal sequence in S2.5, calculate the smallest positive integer n such that 2 n ≥l, fill in 0 from the end of the signal to make the signal length 2 n , and then up-sampled, and after up-sampling, passed through a bandpass filter to improve the calculation efficiency and avoid spectrum leakage of the signal.

4. According to claim 3, a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device is characterized in that: In step S2.1, the blank paragraph length t0 depends on the performance of the device and the temperature measurement environment.

5. According to claim 1, a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device is characterized in that: In the steps S3-S5, assuming that the difference in acoustic path length is d, and the time delays of the dual-channel signal and the excitation signal are τ1 and τ2 respectively, the ambient sound speed is:

6. According to claim 5, a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device is characterized in that: The average ambient temperature of the sound path is obtained by the following formula: Among them, T0 = 273.15K, c0 = 331.5m / s is the sound speed corresponding to T0, and the ambient temperature can be obtained by substituting the sound speed in formula (1) into formula (2).

7. According to claim 1, a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device is characterized in that: In the initialization of step S6, three temperature measurement scenarios are considered, namely, the device is placed horizontally on the desktop, held vertically with one hand, and held horizontally with both hands, and different processing strategies are adopted for different scenarios.

8. According to claim 7, a multi-scenario correction method for measuring room temperature based on a dual-channel microphone intelligent device is characterized in that: For the horizontal desktop scene, after a temperature measurement, according to the external input temperature reference value T R To infer the reference speed of sound c R , and then calculate the corrected parameter d′ based on the delay τ2-τ1 calculated during the temperature measurement process. The subsequent temperature measurement process uses d ′ That's it.

9. The multi-scenario correction method for measuring room temperature based on a dual-channel microphone smart device according to claim 8, characterized in that: For the one-handed vertical holding scenario, first use the same method as the desktop horizontal holding scenario to initialize. In addition, the shaking of the user's arm needs to be considered. The Kalman filter algorithm is used to smooth the data to improve the accuracy of temperature estimation.

10. The multi-scenario correction method for measuring room temperature based on a dual-channel microphone smart device according to claim 9, characterized in that: For the two-handed horizontal holding scenario, first use the same method as the desktop horizontal holding scenario to initialize, then use the Kalman filter algorithm to improve the temperature estimation accuracy, and then make the following fine-tuning on the measurement results: in is the corrected temperature, This is the measurement result before correction.

Images