JPH0549580A - Electric cleaner - Google Patents

Abstract

PURPOSE:To eliminate damages to tatami mats with erroneous recognition by sampling a signal from a sensor at a suction port to set first judgment results according to the second judgment results of the condition of a floor surface when the results are outputted. CONSTITUTION:A sensor signal from a floor surface sensor 22 comprising a pressure-sensitive electroconductive rubber is sampled and with a microcomputer 36, a pattern recognition of the sensor signal is performed by a neural network processing to obtain three kinds of judgment results--tatami mat, carpets short and long in piles. As for the tatami mat, a power brush 14 is turned OFF, as for the carpet short in piles, it is rotated at a low speed and as for the carpet long in piles, it is rotated at a high speed. When results of pattern recognition are given, in the carpets, the results of judgment are converted to quantity forcibly according to the number of sampling data, for example, zero in the state of the sensor signal, and time between valleys and a width of crest of the sensor signal. Thus, the power brush is not allowed to rotate on the tatami mat and be damaged.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an electric vacuum cleaner, and more particularly to an electric vacuum cleaner having a suction surface provided with a floor surface sensor and controlling the rotation of a power brush according to the output of the floor surface sensor.

[0002]

2. Description of the Related Art The first prior art of this type of vacuum cleaner was published on May 21, 1990, in Japanese Patent Application No. Hei2-1317.
No. 27 [A47L 9/04].
In this first conventional technique, a suction force sensor is provided on the main body,
According to the output, it is detected whether the floor surface is a floor, a tatami mat, or another floor. Then, the power brush is turned off on the floor or tatami mat, and the power brush is turned on at other times.

A second prior art is disclosed in Japanese Patent Application No. 2-52624 (A47L 9) published on February 22, 1990.
/ 04]. In this second conventional technique, two sets of photo interrupters are provided at the suction port, one of which is turned on when the suction port contacts the floor surface, and the other of which is turned on when the floor surface is a carpet. Then, the power brush is rotated only when both the two photo interrupters are turned on.

The third conventional technique is to constantly rotate the power brush and detect the difference in the load on the power brush caused by the difference in the floor surface with a current sensor to determine the type of floor surface. The fourth conventional technique is to provide a pressure sensor made of a pressure-sensitive conductive rubber at the suction port, and turn off the power brush on the floor or tatami mat and turn on the power brush on the carpet in accordance with the output of the sensor.

[0005]

SUMMARY OF THE INVENTION In the first prior art,
The fact that the suction force varies depending on the floor surface is used, but the suction force sensor is easily affected by swaying of the suction hose, etc., so that there is a problem that the floor or tatami or other discrimination becomes inaccurate. is there. In the second conventional technique, the up / down roller that acts on the photo interrupter for detecting the carpet may not operate properly due to dust clogging, or the photo interrupter may not be accurately detected when dust clogging occurs. There is a problem that the vacuum cleaner is easily affected by unavoidable dust.

In the third prior art, the power brush is constantly rotated, so that the floor surface, especially the tatami mat, is damaged. In the fourth conventional technique, since the pressure-sensitive conductive rubber is used as the pressure sensor, the instability as in the first or second conventional technique is eliminated, but like the first to third conventional techniques. On, turn off the power brush on the floor or tatami,
The problem is that the carpet is damaged, since only two modes of control can be achieved: just turn on the carpet.

[0007] That is, there are generally two types of carpets, one with short hair and one with long hair. In the former case, the rotation speed of the power brush may be small, but in the latter case, it is necessary to scrape up dust sufficiently. The rotation speed of the power brush must be increased. However, in the case of a carpet in any of the conventional techniques, the number of revolutions is set to a uniform value regardless of the type of carpet. Therefore, in the case of a carpet with long naps, it is not possible to sufficiently scrape up dust or a carpet with short naps. In that case, there was a problem that the hair of the carpet would come off.

Therefore, the applicant of the present invention recognizes the signal from the floor surface sensor by pattern to determine whether the floor surface is a floor, a tatami mat, a carpet with short naps or a carpet with long naps. We proposed a new vacuum cleaner that can set the optimum power brush rotation speed (OFF, low speed rotation or high speed rotation) according to the type. In this proposed electric vacuum cleaner, although damage and insufficient lifting of the carpet are eliminated compared with the conventional technology,
I had to make three subtly different judgments, and there was a possibility of erroneous recognition. For example, if the carpet is actually a tatami mat and is mistaken for a long-haired carpet, the power brush is rotated at the highest speed, which may damage the tatami mat. In addition, since there are more types of judgment results than in the conventional case, the power brush may repeat on / off, or may repeat low speed rotation and high speed rotation. Such a state gives anxiety to the user.

Therefore, a main object of the present invention is to provide a vacuum cleaner which does not damage the tatami mat due to misrecognition. Another object of the present invention is to provide an electric vacuum cleaner that does not disturb the user.

[0010]

Briefly, a first invention is an electric vacuum cleaner in which a brush provided in a suction port is rotated by a motor. A sensor that outputs a different signal depending on the state, a sampling means that samples the signal from the sensor, a pattern that receives the sampling data from the sampling means, determines the state of the floor surface, and outputs the first or second determination result. When the recognition unit and the pattern recognition unit output the second determination result, the predetermined setting condition is determined based on the sampling data, and the first determination result is forcibly set according to the result. It is an electric vacuum cleaner equipped with.

A second invention is an electric vacuum cleaner in which a brush provided at a suction port is rotated by a motor, and a sensor which is provided at the suction port and outputs a different signal depending on a surface state of a floor surface, Sampling means for sampling the signal from the sensor, receiving the sampling data from the sampling means to determine the state of the floor surface, and the first or second
The electric vacuum cleaner is provided with a pattern recognition unit that outputs the determination result and a determination unit that determines the determination result only when a predetermined number of the same determination results are output from the pattern recognition unit.

[0012]

The suction port is preferably provided with a swing member which comes into contact with the floor surface and a floor surface sensor made of a pressure-sensitive conductive rubber which receives a pressure according to the swing. The floor sensor, that is, the pressure-sensitive conductive rubber, outputs signals showing different changes according to the difference in floor surface (floor, tatami mat, carpet with short hair or carpet with long hair). This signal is sampled by the sampling means and provided to the pattern recognition means. The pattern recognition means calculates at least the sampling data according to, for example, a neural network process, or extracts a characteristic parameter pattern and compares it with a standard pattern, and outputs at least two, preferably three or more, discrimination results. To do. And the first
The power brush is turned off according to the determination result (for example, tatami), the power brush is rotated at a relatively low speed according to the second determination result (for example, carpet with short bristles), and the third determination result (for example, hair). Rotate the power brush at a relatively high speed according to the long-legged carpet.

In the first invention, the forced setting means is
On the basis of the sampling data, for example, at least one of the number of “0” data among the predetermined number of sampling data, for example, the time from valley to valley of the sensor signal and the width of the peak of the sensor signal, is determined, When it is a constant condition, the tatami mat discrimination result is forcibly set. Second
In the invention, the determining means determines the determination result only when the pattern recognition means obtains a certain number or more of the same results. However, the above-mentioned fixed number may be changed according to the floor, tatami mat, carpet with short naps, or carpet with long naps.

[0014]

According to the first aspect of the present invention, the force setting means forces the determination result to the tatami mat even if the tatami mat is actually a tatami mat, but the power brush rotates and damages the tatami mat. There is nothing to do. Further, in the second invention, the state does not shift unless the same determination result continues for a certain number or more. Therefore, the power brush is repeatedly turned on / off frequently, or the rotation speed thereof is frequently changed. The feeling can be avoided.

If, as in the embodiment, three or more types of discrimination results are output so that the number of revolutions can be optimally set according to the type of carpet, it is possible to achieve the above-mentioned conventional techniques. In comparison with the case where only a uniform number of rotations can be set in the case of a carpet as described above, in the case of a carpet with long naps, the force for scraping dust is insufficient, or in the case of a carpet with short naps You can surely eliminate defects such as falling out.

The above-mentioned objects, other objects, features and advantages of the present invention will become more apparent from the following detailed description of the embodiments with reference to the drawings.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The suction port 10 of the electric vacuum cleaner shown in FIG. 1 is connected to a main body (not shown) by a hose 12. The suction port 10 is provided with a power brush 14, and the power brush 14 is provided with a belt 16 for the power brush motor 1.
The driving force from 8 is received. The power brush motor 18
For example, a DC motor. In the suction port 10,
A floor sensor 20 is provided. As can be seen from FIG. 2, the floor surface sensor 20 includes a housing 22 having an opening on the lower surface, and an oscillating member 24 is arranged in the opening portion of the housing 22. A coil spring 28 is inserted between the swing member and a contact piece 26 supported by the housing 22 so as to be vertically movable. A pressure-sensitive conductive rubber 32 is attached by a holder 30 to the upper surface of the inner surface of the housing 22 facing the contact piece 26, that is, the swing member 24. The pressure-sensitive conductive rubber 32 changes its resistance value in accordance with the pressure received from the contact piece, for example, so that it decreases as the pressure increases.

As shown in FIG. 3, a DC voltage of, for example, 5V is applied to the floor surface sensor 20 including the pressure-sensitive conductive rubber 32. That is, a DC voltage of 5V is applied to the pressure-sensitive conductive rubber 32.
Is divided according to the resistance value of the resistor 34 and the resistance value of the resistor 34 provided separately, and the voltage is input to the microcomputer 36. However, since the microcomputer 36 is generally provided in the user's hand, the floor sensor 20 and the DC power supply and the microcomputer 36 are connected by two wires via, for example, a hose or a pipe.

When the suction port 10 (FIGS. 1 and 2) is moved in contact with the floor surface, a sensor signal having a pattern as shown in FIG. 4 is input from the floor surface sensor 22, that is, the pressure-sensitive conductive rubber 32. It That is, in the case of the floor, the rocking piece 24 that constitutes the floor surface sensor 20 does not contact the floor surface 38 (FIG. 2), so the rocking piece 24 does not move up and down, and therefore the contact piece 26 does not feel. The piezoelectric rubber 32 is not pushed. Therefore, in the case of the floor, no significant sensor signal is input to the microcomputer 36 as shown in FIG. Further, in the case of a tatami mat, the rocking piece 24 finely rocks according to the tatami mat, and therefore the contact piece 26 frequently repeats contact and separation with the pressure-sensitive conductive rubber 32, as shown in FIG. 4 (B). A sensor signal whose level greatly changes in a short time can be obtained. Further, in the case of a carpet having short naps (hereinafter, may be simply referred to as “cut”), the swing piece 24 of the floor surface sensor 20 is pushed up against the coil spring 28,
Since the surface is relatively smooth, the vertical movement of the rocking member 24 is gentle, and the floor sensor 20 or the pressure-sensitive conductive rubber 32 indicates a relatively large level for a long time as shown in FIG. 4 (C). A sensor signal is obtained that lasts over. Furthermore, in the case of a long-haired carpet (hereinafter sometimes simply referred to as "shaggy"), it does not change as frequently as in the case of tatami, but it changes considerably more frequently than in the case of cutting. The sensor signal as shown in FIG.

In response to the sensor signal from the floor surface sensor 20, the microcomputer 36 performs the processing shown in FIG. However, in FIG. 4, the area indicated by "X" shows the sensor signal when the suction port is pushed and the area shown by "Y" shows the sensor signal when the suction port is pulled. Is used, the sensor signal in the region X where is clearer is used.

Such a sensor signal from the floor sensor 20 is given to the microcomputer 36, and the microcomputer 36 converts the sensor signal into sensor data by using the built-in A / D converter. .. Then, in the preprocessing, for example, the "rising edge" of the sensor signal is detected based on the sensor data, and then the neural network processing (neuro operation) is executed for pattern recognition. The neural network shown in FIG. 6 has 10 input layers, 3 intermediate layers and 1 input layer.
Each of the output layers is configured to perform threshold processing between the intermediate layer and the output layer, and the output layer is similarly subjected to threshold processing. As described above, in this embodiment, the threshold processing is used in order to save the number of intermediate layers and output layers, that is, the calculation time, but it goes without saying that this may be configured as a full network. Then, in each layer of the neural network, as is well known, the equation Y = Σ (XiWj)
According to the above, the input value Xi is multiplied by a predetermined weighting factor Wj, and the output value Y is obtained by the sum of them. However, as the weighting factor Wj, the optimum value obtained by the experiment is set in the ROM. By such neural network processing, it is determined which of the sensor signals shown in FIG. 4 the sensor signal input at that time corresponds to, and three recognition results, that is, tatami (R1), cut (R2) or Shaggy (R3) is obtained.

According to the judgment result obtained by the neural network processing in this way, the microcomputer 36 is driven by the motor driver 40 shown in FIG. 3 or 5.
Is controlled to control the rotation of the power brush motor 18, that is, the power brush 14 (FIG. 1). That is, when the determination result is the tatami (R1) or the floor, the microcomputer 36 stops the power brush motor 16 and does not rotate the power brush 14. In the case of the cut (R2), the microcomputer 36 uses, for example, a PW.
An appropriate control signal, such as the M signal, is applied to the motor driver 40 so that the power brush motor 16 or the power brush 1
Rotate 4 at a relatively low speed. On the other hand, in the case of shaggy (R3), the microcomputer 36 controls the motor driver 40 so as to rotate the power brush 14 at a relatively high speed. In the case of the floor, since a significant sensor signal cannot be obtained as shown in FIG. 4, it can be easily detected by the method described later without using neural network processing.

In this way, the rotation of the power brush 14 is turned off in the case of a floor or tatami, and the power brush rotates in a low speed in the case of a carpet with short fluff to prevent the damage.
For long-haired carpets, rotate the power brush at high speed so that the dust can be sufficiently scraped. Therefore, since the optimum number of rotations of the power brush 14 is set according to the type of carpet, there is a conflict between "damage to the carpet" and "sufficient scraping of dust" that cannot be solved by any of the above-mentioned conventional techniques. Can meet the requirements of both.

Next, another embodiment of the present invention will be described with reference to FIGS. In step S1 shown in FIG. 7, the sensor signal from the floor surface sensor 20 is sampled every 2 milliseconds. In step S2, it is determined based on the sampling data whether or not there is a rising edge. That is, it is determined that the sensor signal has risen when it rises by a predetermined value or more as compared with the previous sampling data. Then, if the rising edge is not detected in step S2, it is determined in the next step S3 whether or not a fixed time, for example, 20 milliseconds has elapsed. If the rise of the sensor signal is detected within this fixed time, that is, if "NO" is determined in step S3, the process shifts to the neural network process described above. However, if it is determined that the sensor signal does not rise even after a lapse of a certain period of time, that is, if "YES" is determined in step S3, the limiter flag is set in step S4 and the subsequent steps are executed.

In this embodiment, as a pre-processing of the neural network processing, in step S5, 10 or a fixed time (20
10 milliseconds (every 2 milliseconds) from the timing when a millisecond has elapsed
Add all the sampling data of. Then, in step S6, it is determined whether the addition result is "1" or less. In steps S5 and S6, as shown in FIG.
It is determined whether or not the zero level of the sensor signal shown in is continuous. In the case of the floor shown in FIG. 4A, the sampling data are all "0". Therefore, in the case of the floor, "YES" is detected in step S6. Therefore, after the skip flag indicating that the neural network processing is not passed is set in step S7, the floor is set as the floor surface determination result in step S8. In this way, it can be easily identified that the floor surface is the floor.

However, if "NO" is determined in step S6, that is, if the added value is "1" or more, it is determined in the next step S9 whether the limiter flag is set. The limiter flag is set when the rising edge is not detected within the fixed time as described above. In this case, the skip flag is set in step S10. If the added value is "1" or more and the limiter flag is not set, the neural network processing as described above is performed in the next step S11.

Then, it is judged in step S12 whether the result of the neural network processing is cut (R2) or shaggy (R3). When a cut or shaggy judgment result is obtained in the neural network processing, in the next step S13, the number of "0" data among the 10 sampling data in the previous step S5 is counted. Then, in step S14, it is determined whether or not the count value C is "5" or more. If there are five or more “0” data in the 10 sampling data, the floor surface determination result is set to tatami in step S15 regardless of the result of the neural network processing in step S11. As shown in FIG. 4B, when the floor surface is a tatami mat, the rocking member 24 (FIG. 2) frequently moves up and down due to the effect of the tatami mat. The number of times sampling data is obtained is large. Therefore, if "YES" is determined in step S14, the floor surface determination result is forced to tatami even if it is determined in step S11 that it is a cut or shaggy. In step S16, it is determined whether the time T from the valley of the sensor signal (FIG. 4) to the valley is 3 seconds or more based on the previous sampling data. If this time is 3 seconds or more, the previous step is performed. Proceed to S15. That is, FIG. 4C or FIG.
In the case of the cut or shaggy shown in (D), the sensor signal does not change as frequently as in the case of the tatami mat, so the time T from the valley to the valley continues for 3 seconds or more. For the same reason, in step S17, the width TM of the peak of the sensor signal (the peak when the sensor signal is other than “0”) is 2
It is determined whether it is in the range of 00 milliseconds to 400 milliseconds. In this way, the determination result by the neural network processing is forcibly set to tatami in step S15 according to at least one condition of steps S14, S16, and S17. When the floor surface is a tatami mat, it is necessary to stop the rotation of the power brush 14 (FIG. 1) without fail. Therefore, in the neural network processing, it was mistakenly recognized as a cut or shaggy despite the tatami mat. This is to make it function as a fail safe in the case.

After step S15 or step S17
After determining "NO" in step S8, the number of times the same determination result continues is counted in step S18 as in step S8 shown in FIG. Then, the judgment result is determined according to Table 1 based on the count value. This table 1
Indicates the number of times required to shift from the current state shown on the left side to the state shown on the upper side.

[0029]

[Table 1]

For example, in the case where the current state is carpet 1 (cut), if the determination result indicating that the carpet is a tatami mat continues for 10 times, the carpet is transferred to the tatami mat. On the other hand, when it is a tatami mat now, carpet 2
If the same judgment result indicating (shaggy) does not continue 40 times, the carpet 2 (shaggy) does not move. That is, the number of transitions from cut or shaggy to tatami is set small, and the number of transitions from tatami to cut or shaggy is set large. As mentioned above, this is a condition that the power brush should not rotate as much as possible in the case of tatami, but it is a problem even if the power brush rotates with some delay in the case of carpet. It will not happen. When the result is confirmed, the microcomputer 36 controls the motor driver 40 and the rotation of the power brush motor 16, that is, the power brush 14 (FIG. 1) in accordance with the determination result in the same manner as described above. .. Also,
In this way, if the state is not shifted unless the same result continues for a certain number of times or more, frequent repetition of on / off of the power brush and frequent changes of its rotational speed are eliminated.

In the above-mentioned embodiment, the neural network processing is executed as the pattern recognition means. However, a well-known pattern recognition method of extracting the characteristic parameter pattern and comparing it with the standard pattern may be used. For example, there is a method based on the number of zero crossings. For example, a certain voltage level indicated by “Vref” in FIG. 4 is set as a threshold value, and the number of times (zero crossing number) the threshold value crosses within a fixed time is extracted as a characteristic parameter. In the case of the floor shown in FIG. 4 (A), the number of zero crossings is “0”. In addition, the number of zero-crossings in the case of tatami is the largest, and the number of zero-crossings in the case of carpet is smaller than that of tatami mats but larger than that of cuts. You may make it output 3 or more types of recognition results by such a number of zero crossings.

Further, other than the extraction of the number of zero crossings as the characteristic parameter, other generally used characteristic parameter may be used. Examples include regression analysis methods, fast Fourier transforms, maximum and minimum methods, etc. In addition, in the above-described embodiment, the microcomputer 36 is configured to obtain three types of determination results. However, the present invention can be similarly applied to, for example, the fourth conventional technique, which can obtain only two types of determination results.

[Brief description of drawings]

FIG. 1 is an illustrative view showing a suction port to which the present invention is applied.

FIG. 2 is an illustrative view showing a cross-sectional structure of the suction port of FIG.

FIG. 3 is a block diagram showing an embodiment of the present invention.

FIG. 4 is a waveform diagram showing a sensor signal.

5 is a functional block diagram showing functions of a microcomputer in the embodiment of FIG.

6 is an illustrative view showing the neural network shown in FIG. 5. FIG.

FIG. 7 is a flowchart showing a part of the operation of another embodiment of the present invention.

FIG. 8 is a flowchart showing another portion of another embodiment of the present invention.

[Explanation of symbols]

 10 ... Suction port 14 ... Power brush 18 ... Power brush motor 20 ... Floor sensor 36 ... Microcomputer 40 ... Motor driver

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] September 11, 1991

[Procedure Amendment 1]

[Document name to be corrected] Drawing

[Correction target item name] All drawings

[Correction method] Change

[Correction content]

[Figure 1]

[Fig. 2]

[Figure 3]

[Figure 4]

[Figure 5]

[Figure 6]

[Figure 7]

[Figure 8]

Claims (2)

[Claims] 1. An electric vacuum cleaner in which a brush provided at a suction port is rotated by a motor, the sensor being provided at the suction port and outputting a different signal in accordance with a surface state of a floor surface. Means for sampling the signal of the above, a pattern recognition means for receiving the sampling data from the sampling means and for judging the state of the floor surface and outputting the first or second judgment result, and the pattern recognition means for the second The vacuum cleaner is provided with a compulsory setting unit that judges a predetermined condition based on the sampling data and compulsorily sets the first judgment result in accordance with the result when the judgment result is output. 2. An electric vacuum cleaner in which a brush provided at a suction port is rotated by a motor, wherein the sensor is provided at the suction port and outputs a different signal depending on a surface condition of a floor surface. Sampling means for sampling the signal of, the pattern recognition means for receiving the sampling data from the sampling means and judging the state of the floor surface and outputting the first or second judgment result, and a predetermined number of the pattern recognition means. An electric vacuum cleaner comprising a confirmation unit that confirms the determination result only when the same determination result is output.