Electric toothbrush oral positioning method based on earphone reverse signal

Abstract

The electric toothbrush oral positioning method based on earphone reverse signal uses three devices of an electric toothbrush, an earphone, a sound card or a smart phone; the earphone is connected with a microphone interface of the sound card; the steps include signal timing cutting; left and right tooth areas are judged by using left and right sound channel power; a roughness algorithm is designed to identify upper and lower jaws by using natural frequency difference, including natural frequency band feature extraction, discriminant threshold calculation based on the natural frequency band feature and tooth area judgment based on the discriminant threshold. The application acquires sensing data based on earphone reverse signal, does not need additional complex devices, the device is cheap, the recognition precision is high, the calculation is fast, and the cost is small. The physical model constructed by the application can directly solve parameters by using a group of tooth brushing data without training, a new tooth position identification scheme is provided for a tooth brushing monitoring system, the user tooth brushing process can be effectively supervised, and it is important to improve the oral health level of the user.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of sound wave perception of human-computer interaction, and particularly relates to an oral positioning method of an electric toothbrush based on a reverse signal of a headset. BACKGROUND

[0002] Research shows that the brushing duration of an electric toothbrush plays an important role in improving the oral health. A sufficient brushing duration can effectively remove plaque bacteria, so that the oral health of a user can be kept at a good level for a long time.

[0003] Designing an accurate and effective brushing position monitoring system to remind and record the brushing of a user can help improve the brushing completion of the user, and thus improve the oral health of the user.

[0004] However, there is no method capable of automatically judging the position of an electric toothbrush in an oral cavity in the prior art. Only by first solving the way of signal collection and then designing a scheme for identifying the collected signal, can the positioning of the electric toothbrush in the oral cavity be realized, and thus the brushing duration of each tooth can be counted to provide reliable data for oral health. SUMMARY

[0005] The present application aims to overcome the deficiencies of the prior art, and provides an oral positioning method of an electric toothbrush based on a reverse signal of a headset, which realizes the identification of the position of the electric toothbrush in an oral cavity.

[0006] The technical concept of the present application is to establish a sound attenuation model of a skull to cochlea bone conduction path, to position the current brushing position by calculating the sound attenuation degree, and thus to realize the brushing duration counting of each tooth on the basis of the brushing position identification function. After each brushing, the system can remind the user of the positions of missed brushing and insufficient brushing duration, so as to urge the user to perform complete tooth cleaning.

[0007] The oral positioning method of the electric toothbrush based on the reverse signal of the headset is characterized in that the method uses three devices of an electric toothbrush, a headset, a sound card or a smart phone; the headset is connected to the microphone interface of the sound card, or connected to the 3.5mm interface of the smart phone; the toothbrush is a separate device and is not connected to other devices;

[0008] The method comprises the following steps:

[0009] 1) The gum is divided into four tooth regions of upper left, upper right, lower left and lower right with the interdental space of the front teeth as the boundary, the headset is worn first, and then the electric toothbrush is started to begin brushing, the vibration of the electric toothbrush causes the skull to generate bone conduction sound wave signals of the four tooth regions, and the reverse signal is obtained by using the headset as a sensor to perceive the sound wave signals, wherein the collection frequency of the sound card or the smart phone is denoted as fs;

[0010] 2) Conversion and timing cut of reverse signals

[0011] Each group of reverse signals collected is converted into discrete digital signals by AD, cached in real-time sequence form, and obtained as fixed-length sequences by timing cut. The timing cut is to set the cutting window length as n, i.e. truncated when each sound channel receives n sampling points, and then perform the next cutting;

[0012] The discrete digital signals form a 2-d array stored in the form of n x 2 after cutting, and the left channel signal sequence is defined as s l (i), and the right channel signal sequence is defined as s r (i), where i represents the sequence index, i = 0, …, n-1;

[0013] 3) Construction of tooth area discrimination method, including 3.1) left and right tooth area discrimination method and 3.2) upper and lower tooth area discrimination method

[0014] 3.1) Left and right tooth area discrimination method is to compare the power of left and right channels

[0015] The calculation method of the left channel power p l and the right channel power p r of the discrete digital signals in a window is as follows:

[0016]

[0017] When p l > p r , it is determined that the tooth brushing position corresponding to the window is the left tooth area, and when p l < p r , it is determined that the tooth brushing position corresponding to the window is the right tooth area;

[0018] 3.2) Upper and lower tooth area discrimination method is to use the different natural frequencies of the upper and lower jaws to determine

[0019] The vibration frequency generated by the electric toothbrush when working is consistent with the main peak frequency of ERS, denoted as F main , and the frequency band width is set to 40 Hz, and the frequency band interval is called the main frequency band; the natural frequency of the upper jaw is set to 500 Hz, and the natural frequency of the lower jaw is set to 1200 Hz; the frequency band width of the natural frequency is set to 200 Hz, and the frequency band interval is called the natural frequency band;

[0020] The following methods are used to distinguish the upper and lower tooth areas from the reverse signals, including 3.2.1) natural frequency band feature extraction step, 3.2.2) discrimination threshold calculation step based on natural frequency band features, and 3.2.3) tooth area determination step based on discrimination threshold;

[0021] The 3.2.1) natural frequency band feature extraction is to convert the signal sequence s(i) into a sequence f(i) by FFT operation and normalization, which is called a frequency domain sequence. The total length of the sequence is still n, and for the frequency interval [F1, F2], Find the index values i corresponding to the frequency domain closest to F1 and F2 from the sequence f(i), respectively, denoted as i1 and i2. The roughness Ra in the frequency interval [F1, F2] is the normalized second-order difference square sum of the frequency domain within the range of i1 to i2.

[0022] Then calculate the roughness Ra of the high frequency band and the low frequency band excluding the main frequency band h And Ra l ;

[0023] Then calculate the difference RaD between Ra h And Ra l , as the final natural frequency band feature:

[0024] RaD = Ra h -Ra l

[0025] The 3.2.2) discriminant threshold calculation includes calculating the mean value of the RaD values of each group of reverse signals in the upper and lower tooth regions, respectively. The discriminant threshold is the average of the RaD mean values of the upper and lower tooth regions.

[0026] 4) Tooth region discrimination, first judge left and right tooth regions, then judge upper and lower tooth regions

[0027] For a group of reverse signals in the window, first judge the left and right tooth regions according to step 3.1), then calculate the RaD value according to the aforementioned step 3.2.1), and compare it with the threshold value obtained in step 3.2.2). When it is greater than the threshold value, it is determined as the upper tooth region, and when it is less than the threshold value, it is determined as the lower tooth region, so as to determine which one of the four tooth regions the window corresponds to.

[0028] In the step 2), in order to avoid spectral leakage in subsequent calculation (the phenomenon of additional frequency band in the neighborhood of spectral peak value caused by binary operation of computer), the window length n of the time sequence cutting satisfies the following condition: n∈R n , R n is related to the sampling frequency fs:

[0029]

[0030] In the 3.2.1) natural frequency band feature extraction, the roughness Ra in the interval [F1, F2] is calculated as follows:

[0031] Firstly, the signal sequence s(i) is converted into sequence f(i) by FFT operation and normalization, which is called frequency domain sequence, and the total length of the sequence is still n, and there is a fixed step frequency F between adjacent f(i) related to n and fs step According to the Nyquist sampling theorem, the calculation formula is:

[0032]

[0033] Secondly, the function I(F obj ) is established,

[0034] I(F obj ) = arg min i |i·F step -F obj |, i = 0, …, n-1,

[0035] arg min i is the value of i when the latter formula reaches the minimum, and its function is to find the index value i closest to F obj from the sequence f(i) for the given frequency domain F obj ;

[0036] Then the first-order forward difference of the sequence f(i) is calculated:

[0037] Δf(i) = f(i+1) - f(i), i ∈ [1, n-1] ∩ N

[0038] Further, the second-order forward difference of f(i) is calculated:

[0039] Δ 2 f(i) = Δf(i+1) - Δf(i), i ∈ [1, n-2] ∩ N

[0040] Since it is a double sound track, the frequency domain sequences f l / r (i) and f l (i) of the left and right sound tracks s r (i) need to be combined for calculation; for the frequency interval [F1, F2], The roughness Ra in this interval is the normalized square sum of the second-order difference, which is expressed as a function according to the above formula:

[0041]

[0042] Thus, a function for calculating the roughness Ra in a certain frequency interval is established.

[0043] In the 3.2.1) natural frequency band feature extraction, the roughness Ra h and Ral ;

[0044] Two high-frequency natural frequency bands R f_h and low-frequency natural frequency bands R f_l are excluded after the main frequency band

[0045]

[0046] where "-" represents the subtraction operation, and the calculation result is in the form of an interval. According to the different F main values, the calculation result is divided into three cases:

[0047] R f_h and R f_l are continuous frequency bands. If , the high-frequency natural frequency band and the low-frequency natural frequency band excluded after the main frequency band are denoted as R f_h = [f h1 , f h2 ], R f_l = [f l1 , f l2 ], where f h1 and f h2 represent the upper and lower limits of R f_h , and similarly, f l1 , f l2 represent the upper and lower limits of R f_l , R f_h has an interval gap, and R f_l has no interval gap; if F main ∈ (1120, 1280), the high-frequency natural frequency band and the low-frequency natural frequency band excluded after the main frequency band are denoted as R f_h = [f h1 , f h2 ]∪[f h3 , f h4 ], R f_l = [f l1 , f l2 ], [f h1 , f h2 ], and [f h3 , f h4 ] represent two continuous frequency bands of R f_h , which are separated by the main frequency band, R f_h has an interval gap, and R f_l has no interval gap; if F main ∈ (420, 580), the high-frequency natural frequency band and the low-frequency natural frequency band excluded after the main frequency band are denoted as R f_l = [f l1 , f l2 ]∪[f l3 , fl4 ], R f_h = [f h1 , f h2 ], [f l1 , f l2 ] and [f l3 , f l4 ] respectively represent R f_l two continuous frequency bands separated by the main frequency band;

[0048] Different calculation methods are used for Ra h and Ra l in the three cases:

[0049] For case 1, the calculation method for Ra h and Ra l is:

[0050]

[0051] For case 2, F main ∈(1120,1280), the calculation method for Ra h and Ra l is:

[0052]

[0053] For case 3, F main ∈(420,580), the calculation method for Ra h and Ra l is:

[0054]

[0055] The 3.2.2) calculation of the discriminant threshold based on the natural frequency band characteristics is as follows:

[0056] Step 1) The number of collected upper and lower dental area reverse signal groups is respectively denoted as n u and n l Since one RaD value is calculated for each group of reverse signals, the RaD of the xth group in the upper dental area is denoted as RaD u (x), and the RaD of the yth group in the lower dental area is denoted as RaD l The mean value M u / l of each group of signals in the upper and lower dental areas is calculated, and the calculation method is:

[0057]

[0058] The discriminant threshold is the average of the above two mean values, that is:

[0059] TH RaD= (M u + M l ) / 2.

[0060] The earphone reverse signal is collected by using the electromagnetic induction phenomenon of the earphone as a signal collection device, and the bone conduction signal from the skull is collected in reverse. When the vibration energy of the electric toothbrush acts on the teeth, the vibration signal is transmitted along the facial bones to the vicinity of the cochlea, and is captured by the earphone and converted into an electric current signal. In the sound card, the sound card can collect the earphone reverse signal by modifying the driver program and the like.

[0061] With the popularity of electric toothbrushes and the continuous reduction of the cost of household earphones, using the earphone reverse signal to realize the position monitoring of the electric toothbrush is a valuable application development method.

[0062] Due to individual differences in oral conditions and skull shapes, it is difficult to use a single model to identify the tooth brushing position. In the past signal recognition technology, the widely used method is to use a CNN convolutional neural network to train the tooth brushing signal. However, if this scheme is used for electric toothbrush oral positioning, not only a large amount of user tooth brushing data needs to be collected, but also the natural frequency information of the upper and lower teeth is difficult to extract. The present application constructs a physical model of tooth position to sound attenuation power, which contains a plurality of unknown parameters reflecting the individual differences of users. The present application only needs to collect a group of user data to solve the parameters, without the need for a large amount of user data training, greatly improving the usability.

[0063] Benefiting from the identification scheme of the physical model, the present application can significantly improve the processing speed and operating efficiency, and provides an important basis for realizing real-time position recognition. Since the physical parameters in the model reflect individual differences, different parameters of users can be used to realize user identity identification under certain conditions. The present application makes it possible for the earphone to become a user identity identification device.

[0064] Compared with the deep learning scheme, the present application has the advantages of low equipment price, simple operation, fast calculation, small overhead, high identification accuracy, and can be used for daily application in household scenarios, and provides important support for improving the oral health level of Chinese residents. BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1 The overall flowchart of the present application.

[0066] Figure 2 The specific work flowchart of the present application.

[0067] Figure 3 The left and right channel power comparison of the four tooth regions. DETAILED DESCRIPTION

[0068] The application is used for sensing the tooth area position of an electric toothbrush in the oral cavity. The dental bed is first divided into four tooth areas, namely, upper left (UL), upper right (UR), lower left (LL), and lower right (LR). The definition of the four tooth area classification is that all the teeth of the upper dental bed belong to the upper tooth area, and all the teeth of the lower dental bed belong to the lower tooth area; in the upper and lower tooth areas, the tooth gap is taken as the dividing line, and the left side is the left tooth area and the right side is the right tooth area. The earphone reverse signal (ERS) is a method of using an earphone as a sensor to sense the bone conduction sound wave signal in the skull by inserting the earphone into the sound card microphone jack. The application uses this method to sense the vibration signal of the electric toothbrush in the oral cavity. The ERS enters the sound card in the form of an analog signal, is converted into a discrete digital signal through sampling, and is stored in the computer or mobile device in the form of an array (sequence). The sampling rate of the sound card is fs, and the default value of fs is 44100 Hz when the device is running on a general household device. When the signal starts to be collected, continuous ERS will be collected by the sound card, so that continuous real-time sequences are obtained in the computer, and the measurement unit of the sequence is the number of sampling points. Since the left and right earphones each have a collection position, the collected real-time sequences are double sound channels.

[0069] The overall flow of the electric toothbrush oral cavity positioning method based on the earphone reverse signal is shown in Figure 1 , and the specific flow is shown in Figure 2 . The method uses three devices: an earphone, a sound card or a smart phone, and an electric toothbrush; the earphone is connected to the microphone interface of the sound card or the 3.5 mm interface of the smart phone; the toothbrush is a separate device and is not connected to other devices.

[0070] The method includes the following steps:

[0071] 1) When using, the earphone is first worn, and then the system is started to begin brushing teeth. The earphone is used as a sensor to sense the bone conduction sound wave signal in the skull through the earphone reverse signal, and the signal is derived from the vibration of the electric toothbrush in the oral cavity.

[0072] 2) Conversion and time sequence cutting of the reverse signal

[0073] The collected reverse signal is converted into a discrete digital signal through AD conversion and is stored in the form of a real-time sequence. After time sequence cutting, a fixed-length sequence is obtained for subsequent processing. The time sequence cutting includes setting the length of the cutting window to n, i.e., the two sound channels are truncated when receiving n sampling points, and the next cutting is performed. In order to avoid spectral leakage in subsequent calculation (the phenomenon that additional frequency bands appear in the neighborhood of the spectral peak value due to computer binary operation), the following condition needs to be met: n∈R n , where Rn Related to the sampling rate fs:

[0074]

[0075] The digital signal is cut to form a 2-d array stored in the form of n x 2, the left channel array is defined as s l (i), and the right channel array is defined as s r (i), where i represents the sequence index, i = 0, …, n-1;

[0076] 3) Tooth area discrimination, first judge left and right tooth area, then judge upper and lower tooth area

[0077] 3.1) The left and right tooth area discrimination method is to compare the left and right channel power

[0078] The left channel power of a window signal is p l , and the right channel power p r is calculated as follows:

[0079]

[0080] When p l > p r , it is determined that the current brushing position is the left tooth area, and when p l < p r , it is determined that the current brushing position is the right tooth area.

[0081] 3.2) Upper and lower tooth area discrimination, use the different natural frequencies of the upper and lower jaws to judge the upper and lower tooth area

[0082] The vibration frequency generated by the electric toothbrush when working is consistent with the ERS main peak frequency, denoted as F main , and the frequency band width is set to 40 Hz, and the frequency band interval is defined as the main frequency band; the natural frequency of the upper jaw is set to 500 Hz, and the natural frequency of the lower jaw is set to 1200 Hz; the natural frequency band width is set to 200 Hz, and the frequency band interval is defined as the natural frequency band; the actually collected ERS contains the above three kinds of frequency band components, among which the natural frequency band is the key information for judging the upper and lower tooth area, and the main frequency band is interference information; for different toothbrushes, the F main value is different, and the main frequency band will intersect with the natural frequency band, causing interference in the judgment.

[0083] The following algorithm is used to discriminate the upper and lower tooth area from the above mixed frequency band information; specifically including a feature extraction method that can exclude the main frequency band (natural frequency band feature extraction), a discrimination threshold calculation method based on the former feature, and a tooth area judgment method based on the discrimination threshold;

[0084] a) Natural frequency band feature extraction

[0085] The signal sequence s(i) is converted into a sequence f(i) by FFT operation and normalization, referred to as a frequency domain sequence, and the total length of the sequence is still n. The index of the sequence f(i) does not directly represent the frequency, but is related to the frequency. There is a fixed step frequency F between each index step The step frequency is related to the sequence length n and the sampling rate fs. According to the Nyquist sampling theorem, the specific calculation method is as follows:

[0086]

[0087] Define a function I(F obj ) for obtaining the target frequency F obj corresponding to the index value from the frequency domain sequence f(i). The specific implementation is as follows:

[0088] I(F obj )=arg min i |i·F step -F obj | (4)

[0089] In the signal spectrum, there is a smooth transition band from the peak value to the zero value between the main peak and the two ends of the spectrum, which is referred to as a transition band below. The relationship between the transition band and the main band is that the main band is a subset of the transition band, and the transition band has a larger range. However, not all information on the transition band can interfere with the natural frequency band. Only the main band can interfere with the natural frequency band. There is also an intersection between the transition band and the natural frequency band. The goal of this step is to extract the strength of the natural frequency band, but the existence of the transition band determines that this strength cannot be directly obtained by calculating the power (including the power of the transition band). Instead, the roughness of the curve should be calculated. The following algorithm is used to obtain the power of other frequencies attached to the transition band, such as the power of the natural frequency band. The natural frequency band is attached to the over-curve, which is manifested as a "spur" phenomenon. Therefore, the following concept of frequency domain curve roughness Ra (hereinafter referred to as roughness) is introduced.

[0090] Before calculating the roughness, define the first-order forward difference of the sequence f(i):

[0091] Δf(i)=f(i+1)-f(i),i∈[1,n-1]∩N (5)

[0092] According to formula (5), the second-order forward difference Δ 2 f(i) of f(i) is calculated:

[0093] Δ 2 f(i)=Δf(i+1)-Δf(i),i∈[1,n-2]∩N (6)

[0094] Because there is dual-channel data, it is necessary to merge and calculate the frequency domain sequences of the left and right channels, and the left and right channel s l / r The frequency domain sequences of (i) are f l (i) and f r (i) Now assume there is a frequency interval [F1, F2], where Within this interval, the roughness Ra is a normalized second-order sum of squared differences, which can be expressed as a function by combining formulas (4) and (6):

[0095]

[0096] This function calculates the roughness Ra within a certain frequency range (band).

[0097] The transition band is much larger than the main frequency band; therefore, the following algorithm is used to calculate the roughness after excluding the main frequency band. Furthermore, since there are two natural frequency bands, the roughness Ra of the high-frequency portion (1200Hz) of each natural frequency band needs to be calculated separately. h Roughness Ra in the low-frequency range (500Hz) l The corresponding frequency bands for the two are respectively the high-frequency natural frequency band R. f_h and low-frequency natural frequency band R f_l The main frequency band can be directly excluded from the natural frequency bands. The calculation method for exclusion is as follows:

[0098]

[0099] In formula (8), the "-" between intervals indicates a subtraction operation, and the calculation result is in the form of an interval, according to F main Depending on the value, the calculation result falls into three categories:

[0100] R f_h and R f_l All are continuous frequency bands, provided that The result of formula (8) is defined as R. f_h =[f h1 ,f h2 ], where f h1 and f h2 R respectively f_h The upper and lower limit frequencies, similarly, R f_l =[f l1 ,f l2 ];

[0101] R f_h There are discontinuous intervals, R f_l There is no discontinuous interval, provided that F main ∈(1120,1280); Define R f_h =[f h1 ,f h2]∪[f h3 ,f h4 ] and [f f_l ,f l1 ] respectively represent two segments of frequencies interrupted by R l2 ; and [f h1 ,f h2 ] and [f h3 ,f h4 ] respectively represent two segments of frequencies interrupted by R f_h .

[0102] R f_l exists in the interval, R f_h does not exist in the interval, provided that F main ∈(420,580); define R f_l =[f l1 ,f l2 ]∪[f l3 ,f l4 ], R f_h =[f h1 ,f h2 ]; [f l1 ,f l2 ] and [f l3 ,f l4 ] respectively represent two segments of frequencies interrupted by R f_l .

[0103] Different calculation methods are used for Ra h and Ra l in the three cases:

[0104] For case 1, the calculation method for Ra h and Ra l is:

[0105]

[0106] For case 2, F main ∈(1120,1280), the calculation method for Ra h and Ra l is:

[0107]

[0108] For case 3, F main ∈(420,580), the calculation method for Ra h and Ra l is:

[0109]

[0110] After obtaining the high and low frequency band information, Rah and Ra l The difference between the two, RaD, as the final natural frequency band feature:

[0111] RaD = Ra h -Ra l (12)

[0112] This feature is only a mathematical description of the band information, so the following algorithm uses this feature to distinguish between the upper and lower dental regions;

[0113] b) Discrimination threshold calculation

[0114] In this step, the user first randomly collects a set of data by directly placing the toothbrush against the corresponding dental region to collect data, and the collection duration is not specified, but includes 4 dental regions;

[0115] For data of different dental regions, the RaD values calculated from a) are all different, but there is a statistical rule, which is that the RaD values of the upper dental region are generally greater than those of the lower dental region, and there is a boundary between the RaD values of the upper and lower dental regions, which is called the RaD threshold, denoted as TH RaD ;

[0116] When a user collects a plurality of sets of upper dental region data, the total number of sets is n u , and since each set of data calculates a RaD value, for the upper dental region, the calculation result of the xth set is denoted as RaD u [x],

[0117] Similarly, for the lower dental region, the total number of sets is n l , and the calculation result of the yth set is denoted as RaD l [y],

[0118] First, calculate the RaD mean values M u / l of the upper and lower dental regions respectively, and the calculation method is:

[0119]

[0120] The calculation method of the discrimination threshold is to take the mean value of the above two quantities again, that is:

[0121] TH RaD = (M u + M l ) / 2 (14)

[0122] c) Dental region discrimination

[0123] For a set of window signals, calculate the RaD value according to the previous steps, and compare it with TH RaD , when RaD > TH RaD , it is determined as the upper dental region, and when RaD ≤ THRaD The tooth was initially identified as the lower tooth region; combining the results from section 3.1 above, the final four-tooth region discrimination result was obtained.

[0124] Example 1

[0125] In this embodiment, a 24-year-old volunteer was randomly selected as the system user. This user possessed a complete set of 32 adult teeth. The electric toothbrush used in this embodiment was a Philips HX9911, and the earphone device was an ATH-M30x. The user's operating procedures included data collection, model solving, brushing position recognition, and brushing time statistics. For the system workflow, please refer to [link to documentation]. Figure 1 This embodiment mainly includes the following steps:

[0126] 1) The user holds the toothbrush handle naturally with all five fingers, and the toothbrush head is clamped on both sides of the earphone. First, brushing data is collected.

[0127] 2) Timing Segmentation

[0128] The system receives real-time sequences and performs segmentation, with each channel receiving 4410 sample points as a window. The sampled data is stored as a 2-d array of 4410×2.

[0129] 3) Dental region discrimination

[0130] a) First, determine the left and right dental regions. There is a difference in power between the left and right channels when the user brushes their teeth; the left and right dental regions are determined based on the power levels of the left and right channels. The left channel power p in the current window... l =28425.37, right channel power is p r =4784.85, the system is based on p l >p r The result indicates that the current location is the left tooth region.

[0131] b) Based on the calculation results of the RaD value, such as Figure 3 As shown, if the RaD value of the current brushing position is greater than 99.2, it is determined to be the upper teeth area; otherwise, it is determined to be the lower teeth area. In the current window, RaD = 177.35, therefore the system determines it to be the upper teeth area.

[0132] Based on the above judgment results, the current brushing action is determined to be in the upper left teeth area.

Claims

1. A method for oral positioning of an electric toothbrush based on the inverse signal of a headset, characterized in that The method uses three devices of an electric toothbrush, earphones, a sound card or a smart phone; the earphones are connected with a microphone interface of the sound card or a 3.5mm interface of the smart phone; the toothbrush is a separate device and is not connected with other devices; The method comprises the following steps: 1) dividing the gums into four tooth areas of upper left, upper right, lower left and lower right with the interdental space of the front teeth as the boundary, wearing the earphones first and then starting the electric toothbrush to begin brushing teeth, the vibration of the electric toothbrush generates four bone conduction sound wave signals of the tooth areas, and the earphones are used as sensors to obtain reverse signals to perceive the sound wave signals, wherein the collection frequency of the sound card or the smart phone is denoted as fs; 2) conversion of the reverse signals and time sequence cutting Each group of collected reverse signals is converted into discrete digital signals by AD conversion, cached in real-time sequence form, cut by time sequence to obtain fixed-length sequences; the time sequence cutting is to set the cutting window length as n, i.e. cutting when each sound channel receives n sampling points, and then performing the next cutting; The discrete digital signal is cut to form a 2-d array stored in the form of n x 2, and the left channel signal sequence is defined as s l (i), and the right channel signal sequence is defined as s r (i), wherein i represents a sequence index, i = 0, …, n-1. 3) construction of the tooth area discrimination method, comprising 3.1) left and right tooth area discrimination method and 3.2) upper and lower tooth area discrimination method 3.1) the left and right tooth area discrimination method is to compare the powers of the left and right sound channels the left channel power p of a discrete digital signal of a window l the right channel power p of a discrete digital signal of a window r is calculated as when p l r the tooth brushing position corresponding to the window is the left tooth area when p l r the tooth brushing position corresponding to the window is the right tooth area;​​ 3.2) the upper and lower tooth area discrimination method is to use the different natural frequencies of the upper and lower jaws to make the judgment The vibration frequency generated by the electric toothbrush during operation is consistent with the main peak frequency of the ERS, denoted as F main , the frequency band width is set to 40 Hz, and the frequency band interval is referred to as the main frequency band; the natural frequency of the maxilla is set to 500 Hz, and the natural frequency of the mandible is set to 1200 Hz; the frequency band width of the natural frequency is set to 200 Hz, and the frequency band interval is referred to as the natural frequency band; The following method is used to discriminate the upper and lower tooth areas from the reverse signals, comprising 3.2.1) natural frequency band feature extraction step, 3.2.2) discrimination threshold value calculation step based on the natural frequency band features, and 3.2.3) tooth area judgment step based on the discrimination threshold value; The 3.2.1) natural frequency band feature extraction is to convert the signal sequence s(i) into a sequence f(i) by FFT operation and normalization, which is called a frequency domain sequence, and the total length of the sequence is still n. For the frequency interval [F1, F2], The index values i corresponding to the frequency domains closest to F1 and F2 are found from the sequence f(i), respectively, and are denoted as i1 and i2. The roughness Ra in the frequency interval [F1, F2] is the normalized second-order difference square sum of the frequency domains in the range from i1 to i2. After calculating the roughness Ra of the high frequency band and the low frequency band excluding the main frequency band h and Ra l ; Recalculate Ra h and Ra l The difference RaD between the two as the final natural band feature: RaD= Ra h -Ra l The discrimination threshold value calculation of 3.2.2) comprises calculating the mean values of each group of reverse signals RaD of the upper and lower tooth areas respectively, and the discrimination threshold value is the average of the RaD mean values of the upper and lower tooth areas; 4) tooth area discrimination, first judging the left and right tooth areas and then judging the upper and lower tooth areas For a group of reverse signals in the window, first, according to step 3.1), the left and right tooth areas are judged, then the RaD value is calculated according to the aforementioned step 3.2.1), and compared with the threshold value obtained in step 3.2.2), when greater than the threshold value, it is determined as the upper tooth area, and when less than the threshold value, it is determined as the lower tooth area, so as to judge which one of the four tooth areas the window corresponds to.

2. The earphone reverse signal based electric toothbrush oral positioning method of claim 1, wherein In step 2), to avoid spectral leakage in subsequent calculations, the window length n of the time slicing satisfies the following condition: n ∈ R n , n R is related to the acquisition frequency fs:

3. The earphone reverse signal based electric toothbrush oral positioning method of claim 1, wherein In the natural frequency band feature extraction of 3.2.1), the calculation method of the roughness Ra in the interval [F1, F2] is as follows: First, the signal sequence s(i) is converted into a sequence f(i) by FFT operation and normalization, which is called a frequency domain sequence. The total length of the sequence is still n, and there is a fixed step frequency F between adjacent f(i) related to n and fs step According to the Nyquist sampling theorem, the calculation formula is: Secondly, the function I(F obj ) is established. I(F obj )=arg min i |i·F step -F obj |,i=0,…,n-1, argmin i is the value of i that minimizes the expression obj for a given frequency domain F obj finds the index value i from the sequence f(i) that is closest to F Then, the first-order forward difference of the sequence f(i) is calculated: Δf(i)=f(i+1)-f(n),n∈[1,n-1]∩N Further, the second-order forward difference of f(i) is calculated: Δ 2 f(i) = Δf(i+1) - Δf(i), i ∈ [1, n-2]∩N Since it is a two-channel, it is necessary to calculate the left and right channels s l / r The frequency domain sequence f l (i) and f r (i) are merged to calculate; for the frequency interval [F1, F2], The roughness Ra in this interval is the normalized second-order difference square sum, which is expressed as a function form by combining the above formulas: Thus, a function for calculating the roughness Ra in a certain frequency interval is established.

4. The earphone reverse signal based electric toothbrush oral positioning method of claim 3, wherein In the 3.2.1) natural frequency band feature extraction, the roughness Ra after excluding the main frequency band in the high frequency portion 1200 Hz and the low frequency portion 500 Hz is calculated by the following steps h and Ra l ; Both exclude the high frequency natural band R of the main band f_h and the low frequency natural band R f_l is: where "-" denotes the set difference operation, and the result is in the form of an interval, according to F main Depending on the values of the variables, the result can be in one of three cases: R f_h and R f_l are continuous frequency bands, if then the high and low frequency natural bands after the main frequency band are excluded and denoted as R f_h = [f h1 , f h2 ], R f_l = [f l1 , f l2 ], where f h1 and f h2 represent the upper and lower limits of R f_h , and similarly, f l1 , f l2 represent the upper and lower limits of R f_l ; R f_h has an interval of discontinuity, and R f_l does not have an interval of discontinuity; if F main ∈(1120, 1280), then the high and low frequency natural bands after the main frequency band are excluded and denoted as R f_h = [f h1 , f h2 ]∪[f h3 , f h4 ], R f_l = [f l1 , f l2 ], [f h1 , f h2 ] and [f h3 , f h4 ] represent two continuous frequency bands of R f_h separated by the main frequency band; R f_l has an interval of discontinuity, and R f_h does not have an interval of discontinuity; if F main ∈(420, 580), then the high and low frequency natural bands after the main frequency band are excluded and denoted as R f_l = [f l1 , f l2 ]∪[f l3 , f l4 ], R f_h = [f h1 , f h2 ], [f l1 , f l2 ] and [f l3 , f l4 ] represent two continuous frequency bands of R f_l separated by the main frequency band; Ra for the three cases h Ra for the three cases l Using different calculation methods: For case 1, Ra h and Ra l The calculation method is: For case 2, F main ∈(1120, 1280), Ra h and Ra l The calculation method is: For case 3, F main ∈(420,580), Ra h and Ra l The calculation method is:

5. The earphone reverse signal based electric toothbrush oral positioning method of claim 4, wherein The discrimination threshold value calculation based on the natural frequency band features of 3.2.2) is as follows: Step 1) the number of reverse signal groups of the upper and lower dental regions is respectively recorded as n u , n l Since a RaD value is calculated for each group of reverse signals, the RaD of the xth group of the upper dental region is recorded as RaD u (x), and the RaD of the yth group of the lower dental region is recorded as RaD l (y). The average value M u / l of the RaD of each group of signals of the upper and lower dental regions is respectively calculated, and the calculation method is: The discrimination threshold value is the average of the above two mean values, i.e. TH RaD = (M u + M l ) / 2.

Images