The present invention relates to a cough detecting apparatus and cough detecting method.
Coughing is often observed in the respiratory system disease (e.g., particularly, asthma, chronic obstructive pulmonary disease and bronchitis). At present, diagnosis of coughing depends on a medical examination by interview with a doctor. However, a patient is not always coughing at the time of medical examination. There is no other way for the doctor but to inquire about the subjective symptom of the patient. Further, although the patient is conscious of the symptom in the daytime when he is awake, details of coughing experienced during the sleep cannot be identified. The information that can be obtained will be limited to such a statement that the coughing was severe or that the patient could not sleep due to severe coughing. Thus, objective evaluation cannot be conducted, with the result that effective medical treatment cannot be provided, according to the conventional method.
To conduct objective evaluation of coughing, the apparatuses are disclosed wherein voice signal information characteristic of coughing is stored in advance, and based on the voice signal information characteristic of coughing, a voice signal resulting from coughing is identified and extracted from the voice signal inputted from a microphone and others, whereby coughing is detected and monitored (e.g., Patent Documents 1 through 3).
[Patent Document 1] Unexamined Japanese Patent Application Publication (Tokkaihei) No. 7-376
[Patent Document 2] Unexamined Japanese Patent Application Publication (Tokkaihei) No. 8-38481
[Patent Document 3] Unexamined Japanese Patent Application Publication (Tokkai) No. 2003-38460
However, the apparatus disclosed in the aforementioned Patent Document cannot easily distinguish between coughing and a similar sound when a sound similar to coughing occurs. Such a sound may be detected as coughing. Further, the noise caused by being in contact with a sensor for sampling the voice such as a microphone or piezoelectric device cannot be always distinguished from coughing. When a patient is not coughing, coughing by another person may be captured by the microphone and may be identified as the patient's coughing.
The object of the present invention is to solve the aforementioned problem and to provide a cough detecting apparatus and a cough detecting method capable of detecting coughing with high precision.
The cough detecting apparatus of the present invention includes:
a voice measuring section for detecting subject's voice;
a body motion measuring section for detecting the body motion of the subject; and
a cough detecting section for detecting coughing based on the voice information detected by the aforementioned voice measuring section, and the body motion information detected by the aforementioned body motion measuring section.
The cough detecting method of the present invention includes:
a voice measuring step for measuring subject's voice using a voice measuring section;
a body motion measuring step for measuring the subject's body motion using a body motion measuring section;
a cough detection step wherein the voice information measured by the aforementioned voice measuring step and the body motion information measured by the aforementioned body motion measuring step are inputted into the cough detecting section to detect coughing.
When detecting the sound of coughing, according to the present invention, the body motion (movement of the subject's body) is also detected being associated therewith. A sound similar to coughing having been detected is identified as a non-coughing sound if the body motion inherent to coughing has not been detected. This arrangement ensures high-precision coughing detection.
a) is a diagram representing the voice data Sv(a) and body motion data Sm(a) measured by the cough detecting apparatus of the present embodiment when a subject has coughed.
b) is a diagram showing the voice data Sv(b) and body motion data Sm(b) when another person close to the subject has coughed.
a) is a diagram showing the data on body motion caused by coughing.
b) is a diagram showing the time durations T1, T2, and T3.
c) is a diagram showing the inclination angle θ1 and θ2.
d) is a diagram showing the kurtosis K.
The voice measuring section 10 includes a microphone 11 for converting the inputted voice into a voice signal; an amplifier 12 for amplifying the voice signal converted by the microphone 11; a smoothing circuit 13 for half-wave rectification and smoothing of the voice signal amplified by the amplifier 12; an A/D converter 14 for converting the voice signal processed by the smoothing circuit 13, into voice data; and an I/F 15 for transmitting the voice data having been converted by the A/D converter 14 to the cough detecting section 30.
When sound is measured using the microphone 11 as in the present embodiment, noise will be generated if the human body and clothing are brought into contact with the microphone. This requires the microphone to be fixed not to come in contact with them. However, if the microphone is brought too far away from the human body, there will be an increase in the influence of external noise. Thus, the microphone is preferably placed as close as possible to the human body.
In addition to the microphone, a piezo microphone or accelerometer can be placed in contact with the pharyngeal part or others, and the sound can be detected as vibration. In this case, since the equipment is kept in contact with the human body, it is not much affected by external noise.
The body motion measuring section 20 includes an accelerometer 21 for converting the inputted body motion into a body motion signal; an amplifier 22 for amplifying the body motion signal converted by the accelerometer 21; a filter circuit 23 for removing a low-frequency component below a predetermined frequency and a high-frequency component above a predetermined frequency from the body motion signal amplified by the amplifier 22; an A/D converter 24 for converting the body motion signal processed by the filter circuit 23 into the body motion data; and an I/F 25 for transmitting to the cough detecting section 30 the body motion data having been converted by the A/D converter 24.
The accelerometer 21 for measuring the body motion is preferably located at the abdominal region, chest region or cervical region, more preferably at the abdominal region or chest region. This is because the region around the diaphragm or chest region exhibits a characteristic movement at the time of coughing.
The accelerometer 21 is preferably attached onto the skin directly or indirectly by a double-faced tape coated with adhesive or the like, which is less sensitive to the skin. Further, noise preventive covering material or cushioning material is preferably arranged around the accelerometer 21.
The cough detecting section 30 includes an I/F 31 for receiving the voice data and body motion data from the voice measuring section 10 and body motion measuring section 20; a CPU 32 for processing the voice data and body motion data received by the I/F 31 according to the program; a ROM 33 for storing the program and data required by processing by the CPU 32; a RAM 34 for temporarily storing the program and data required for processing by the CPU 32; an outer memory 35 for storing the result of processing by the CPU 32, into the hard disk, DVD-R or CD-R; an input section 36 for inputting data into the cough detecting section 30; and a display section 37 for displaying the result of processing by the CPU 32.
The cough detecting section 30 may be formed of a special-purpose information processing apparatus or a general-purpose personal computer. In the case of a personal computer, use of an easy-to-carry portable information terminal (PDA) is preferred.
In the present embodiment, each of the voice measuring section 10 and body motion measuring section 20 is provided with an A/D converter 14 and A/D converter 24 respectively. Further, the cough detecting section 30 may be provided with an A/D converter.
In the first place, the CPU 32 takes a decision step to determine whether or not a cough measurement start instruction has been issued by the input section 36 (Step S10). If decision has been made that start of cough measurement has been specified (Step S10: Yes), the CPU 32 initiates the step of storing the voice data and body motion data from the voice measuring section 10 and body motion measuring section 20 into the outer memory 35 via the RAM 34 (Step S11). In this case, voice data and body motion data are stored being associated with the subject ID and the time. If decision has been made that a cough measurement start instruction has not been issued (Step S10: No), the control goes back to the Step S10 to wait until a cough measurement start instruction is given.
Then the CPU 32 takes a decision step to determine whether or not a cough measurement termination instruction has been issued (Step S12). If it has been determined that a cough measurement termination instruction has been issued (Step S12: Yes), the CPU 32 terminates the step of storing the voice data and body motion data from the voice measuring section 10 and body motion measuring section 20 into the outer memory 35 via the RAM 34 (Step S13). If it has been determined that a cough measurement termination instruction has not been issued (Step S12: No), the control goes back to the Step S42 to wait until the cough measurement termination instruction is given.
a) shows the voice data Sv(a) and body motion data Sm(a) measured by a cough detecting apparatus of the present embodiment when the subject has coughed, wherein elapsed time is plotted on the horizontal axis, and data level is plotted on the vertical axis in a schematic representation. Coughing is produced by a sudden exhalation of the air stored in the abdominal region subsequent to closure of the throat. A strong noise is produced in a short period of time, and the abdominal region is dented in a short period of time too. Thus, as illustrated, when coughing occurs, both the rising gradient θv(a) of the voice data Sv(a) and θm(a) of the motion data Sm(a) are steep, and both the time width Tv(a) and Tm(a) are short.
b) is a diagram showing the voice data Sv(b) and body motion data Sm(b) when another person close to the subject has coughed. As illustrated, the rising gradient θv(b) of the voice data Sv(b) is detected similarly to the case of
To be more specific, to determine if the subject has coughed or not, it is effective to check that each of the rising gradient θv of voice data Sv and θm of body motion data Sm is equal to or greater than a predetermined value, and each of the time width Tv and Tm does not exceed a predetermined value.
In the first place, the CPU 32 determines whether or not a cough detection instruction has been inputted from the input section 36 (Step S20). If decision has been made that the cough detection instruction has been inputted (Step S20: Yes), the CPU 32 reads out the voice data corresponding to the subject ID from the outer memory 35, and load it into the RAM 34 (Step S21). If decision has been made that the cough detection instruction has not been inputted (Step S20: No), the control goes back to the Step S20 to wait until the cough detection instruction is inputted.
Then the CPU 32 analyzes a temporal change of the voice data stored in the RAM 34. A decision step is taken to determine whether or not the rising point (tv1 of
Then the CPU 32 analyzes the voice data extracted in the Step S23, and calculates the rising gradient θv of the voice level, time width Tv, and the maximum voice level Svmax (Step S24).
Then the CPU 32 determines whether or not the rising gradient θv of the voice level obtained by calculation and the voice data extracted from the time width Tv are the voice data resulting from coughing. If the rising gradient θv of the voice level is equal to or greater than a predetermined value θv0 and time width Tv is below a predetermined value Tv0, they are determined to be the voice data resulting from coughing. Otherwise, they are determined not be the data resulting from coughing (Step S25).
If it has been determined that the extracted voice data is the data resulting from coughing (Step S25: Yes), the CPU 32 reads the body motion data corresponding to the subject ID from the outer memory 35, and loads it into the RAM 34 (Step S26). If it has been determined that the voice data is not voice data resulting from coughing (Step S25: No), the control jumps to the Step S33.
From the body motion data loaded into the RAM 34 in the Step S26, the CPU 32 extracts the body motion data within a predetermined range determined based on the aforementioned extracted voice data (Step S27). Since during coughing the voice is generated after the movement of the abdominal region and throat region, the body motion is detected earlier than the voice. Thus, the body motion data should be extracted a predetermined time earlier than the rising point tv1 in the voice data. For example, the body motion data are sufficient only when they are extracted from the time which is about 200 msec earlier than the rising point time tv1 of the voice data up to the falling point time tv2 of the voice data.
Then the CPU 32 analyzes the extracted body motion data and determines whether or not the body motion data includes the minimum value Smmin which does not exceed the threshold value Smth (Step S28). If it has been determined that the body motion data includes the minimum value Smmin which does not exceed the threshold value Smth (Step S28: Yes), a step is taken to determine whether or not the data level turns once positive, and then goes to 0 (wherein the time when 0 is reached is represented by tm1 and tm2) when the time is traced, toward both before and after, from the time tmp at which this minimum value Smmin is obtained. To be more specific, a decision step is taken to determine whether or not the waveform is characteristic of coughing (Step S29). The details will be described later.
A step is taken to trace the time toward both before and after from the time tmp at which the minimum value Smmin is obtained. If it has been determined that the data level turns once positive, and then goes to 0 (Step S29: Yes), the CPU 32 analyzes the extracted body motion data, and calculates the rising gradient θm from the minimum value Smmin of the body motion level, the time width Tm from tm1 to tm2, and others (Step S30).
A step is taken to trace the time toward both before and after from the time tmp at which the minimum value Smmin is obtained. If it has been determined that the data level has not turned positive to later become 0 (Step S29: No), the CPU 32 jumps to Step S33.
Based on the rising gradient θm from the minimum value Smmin of the body motion level obtained by calculation in Step S30, and the time width Tm from tm1 to tm2, the CPU 32 determines whether or not the body motion data is the body motion data resulting from coughing. If the rising gradient θm from the minimum value Smmin of the body motion level is equal to or greater than a predetermined value θm0, and the time width Tm from tm1 to tm2 does not exceed a predetermined value Tm0, then the extracted body motion data is determined to be the body motion data resulting from coughing. Otherwise, this data is determined not to be the body motion data resulting from coughing (Step S31).
If the extracted body motion data has been determined to be the body motion data resulting from coughing (Step S31: Yes), the CPU 32 detects the extracted voice data under the conviction that the data results from coughing, and stores the data into the outer memory 35 (Step S32). In this case, the time of occurrence t, the maximum voice level Svmax, the minimum body motion level Smmin and others are stored being associated with the subject ID. Further, the pattern of at least one of voice data and body motion data corresponding to the type of coughing is stored in the ROM 33 or outer memory 35. The type of coughing is identified by comparison with the measured pattern corresponding to the aforementioned pattern, and can be stored being associated with the subject ID. If the extracted body motion data has been determined not to be the body motion data (Step S31: No), the control jumps to Step S33.
In Step S33, the CPU 32 analyzes the temporal change of the voice data after the aforementioned extracted voice data loaded in the RAM 34. The CPU 32 takes a decision step to determine whether or not the rising point tv1 is included in the voice data wherein the voice level has increased from the level below the threshold value level Svth to the level above the threshold value level Svth. If the voice data has been determined to include the rising point tv1 (Step S33: Yes), the control goes back to Step S23, and cough detection is repeated. This means that cough detection has been performed in all the area including the rising point tv1. If the voice data is determined not to include the rising point tv1 (Step S33: No), the flow terminates.
As described above, when detecting coughing, according to the present embodiment, it is detected being associated with the body motion. Even if the voice similar to coughing has been found out, this arrangement determines that it not coughing, if the body motion inherent to coughing has not been detected. Thus, the present embodiment ensures high-precision detection of coughing.
The diagram shows the average value Ave, the minimum value Min, and the median M which is the average value between the average value Ave and the minimum value Min ((the average value Ave+the minimum value Min)/2), which are data levels used in the process of coughing detection to be described later.
Some type of coughing exhibits a waveform wherein there is a rise in the positive direction between P3 and P4, as shown in
In the first place, the average value Ave of the data level of the sampling data in the sampling frame is calculated. In the sampling frame, the sampling point of the minimum value Min (P4 in
If there is a sampling point denoting the minimal value as in
The aforementioned procedure is followed by the step of identifying the sampling point of the maximum value toward the left from the sampling point P3, and sampling point of the maximum value toward the right from the sampling point P4. In
The next step is to identify the sampling point wherein the value has a sudden decrease to reach close to 0 toward the left side from the sampling point P2, and the sampling point wherein the value has a sudden decrease to reach close to 0 toward the right side from the sampling point P5. In
(Another Method of Determining if the Body Motion Waveform Characteristic of Coughing Results from Coughing or not)
Instead of the maximum inclination angle θmax of
At least one of the time durations T1 and T2 of
In the present embodiment, if the voice data was determined as resulting from coughing after analysis of voice data, the body motion data was analyzed to detect coughing. Conversely, if the body motion data is determined to have resulted from coughing by the analysis of the body motion data, coughing can be detected by the analysis of voice data. It is also possible to make such arrangements that voice data and body motion data are analyzed independently of each other. If both are determined to have resulted from coughing, coughing is detected.
Further, coughing can be detected, for example, by the computation such as multiplication of voice data by body motion data. If the data results from coughing, as described above, the portion of the voice data output (portion where voice is detected) and the portion of the body motion data output (portion where body motion is detected) have a predetermined time lag. Accordingly, if the data results from coughing, they are shifted by the predetermined time lag. This will bring about agreement between the portion of voice detection and that of body motion detection. Conversely, if the data does not result from coughing, the portion of voice data output (portion of voice detection) is uncorrelated with that of body motion data output (portion of body motion detection), and there is a high probability that they do not agree with each other.
The aforementioned description suggests that the data subsequent to the computation of multiplication is outputted to be relatively large when the data results from coughing, as compared to the case where the data does not result from coughing. Thus, if the computation of multiplication is performed, whether or not the data results from coughing can be determined by whether or not the output of the data subsequent to computation is larger than a predetermined value.
Number | Date | Country | Kind |
---|---|---|---|
2005-292085 | Oct 2005 | JP | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/JP2006/018108 | 9/13/2006 | WO | 00 | 1/5/2007 |