US3821472A - Cough monitoring apparatus - Google Patents

Cough monitoring apparatus Download PDF

Info

Publication number
US3821472A
US3821472A US00229290A US22929072A US3821472A US 3821472 A US3821472 A US 3821472A US 00229290 A US00229290 A US 00229290A US 22929072 A US22929072 A US 22929072A US 3821472 A US3821472 A US 3821472A
Authority
US
United States
Prior art keywords
feature
signals
occurrence
gate
output
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Lifetime
Application number
US00229290A
Inventor
M Herscher
T Martin
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Threshold Tech
THRESHOLD TECHNOLOGY INC US
Original Assignee
Threshold Tech
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Threshold Tech filed Critical Threshold Tech
Priority to US00229290A priority Critical patent/US3821472A/en
Application granted granted Critical
Publication of US3821472A publication Critical patent/US3821472A/en
Anticipated expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00Speech recognition

Definitions

  • ABSTRACT 2 App 229,290
  • An apparatus that receives ambient sounds and recognizes the occurrence of coughs to the exclusion of other sounds. Means are provided for translating the [52] US. Cl 179/1 SA received sounds into a plurality of Spectra] Component [51] Int. Cl. G] l/04 signals
  • a feature signal generating meanslis provided [58] Field of Search 179/ l 1 1 for sensing the presence of properties of the spectral Vci 128/2 R121 component signals that are characteristic of coughs and for generating a feature signal for each property [56] References C'ted found to be present.
  • An occurrence decision means UNITED STATES PATENTS responsive to the feature signals, generates an occur- 3,198,884 8/1965 Dersch 179/1 SA rence indication p receiving a predetermined cOrn- 3,204,030 8/1965 Olson 179/1 SB bination of feature signals.
  • the occurrence deci i n.means is 3,368,039 2/1968 Clapper 179/1 SA ble to the cough of a particular individual.
  • PATENTEDJUHZB I974 SHEET 5 (IF 5 MANUAL RESET 4 l 2 3 4 l L L L L L 1 3 3 3 E 4 l 2 3 4 5 I D D D D D T A R PE PR R R R W MV MV M AMA A H llll. MV Am A A R R LR LR R LO O O D D V 4 w w m w 2 I I rh F. F. F. F.
  • one prior art device includes a pneumatic band that is fitted around a patients chest and connected to a recorder through pneumatic tubes. Problems of patient discomfort limit this devices usefulness.
  • Another scheme provides a microphone carried on a patients person or positioned nearby. Sounds received by the microphone are transmitted as electrical signals to a receiver which triggers a counter whenever the power of the received signals exceeds a predetermined level.
  • the present invention is directed to an apparatus which receives ambient sounds and reliably recognizes the occurrence of specified non-verbal sounds, e.g., coughs, to the exclusion of other sounds.
  • Means are provided for translating the received sounds into a plurality of spectral component signals.
  • a feature signal generating means is provided for sensing the presence of properties of the spectral component signals that are characteristic of coughs and for generating a feature signal for each property found to be present.
  • An occurrence decision means responsive to the feature signals, generates an occurrence indication upon receiving a predetermined combination of feature signals.
  • the occurrence decision means is programmable for different specified sounds, for example, the cough of a particular individual.
  • FIG. 1 is a graphical representation of the general shape of the energy vs. frequency spectrum for a typical human cough
  • FIG. 2 is a block diagram of an embodiment of a cough monitoring apparatus in accordance with the invention.
  • FIG. 3 is a block diagram of the translating means portion of the invented apparatus
  • FIG. 4 is a block diagram representation that illustrates the functioning of the feature signal generator utilized in the present invention.
  • FIG. 5 is a block diagram of the storage means and occurrence decision logic means utilized in the present invention.
  • FIG. 6 is a side perspective view of a console utilized to house an embodiment of the present invention.
  • FIG. 7 is a block diagram of the duration sensor utilized in the occurrence decision logic means of FIG. 5.
  • coughs have certain combinations of distinguishing characteristics that are usually not present in speech or ambient room noise.
  • a cough is noiselike burst of energy of substantial duration characterized by a sharp onset and having a single broad peak in its energy vs. frequency spectrum as shown in FIG.'1.
  • the frequency components that contain most of the cough energy are in the range about 500-7500 c.p.s., and the broad peak is typically in the range about l,0003,000 c.p.s.
  • Many speech sounds on the other hand, have two or more distinct peaks or formants in their energy vs. frequency. Also, it has been empirically determined that certain slope and energy relationships exist between the different spectral components of most coughs.
  • a tranducer 30 typically a gradient microphone, receives ambient input sounds and produces time-varying electrical signals representative of the received sounds.
  • the sound representative signals are coupled over' a line 30a to a translating means 20 that includes, inter alia, a bank of frequency-selective devices such as bandpass filters.
  • the output of the translating means is a plurality of signals on lines 20a through 20p. These output signals represent spectral components of the received sounds and can be pictured, for example, as reflecting the energy levels at the selected frequencies shown in FIG. 1.
  • a feature signal generating means 50 receives the spectral component signals 20a through 20p and senses the presence of properties of the spectral component signals that correspond to preselected properties or features of coughs. The manner in which this is achieved is described in further detail hereinbelow. It suffices for the present to understand that each of the output lines corresponds to one of the preselected feameans 100 and occurrence decision logic means 150.
  • Storage means 100 is utilized, in conjunction with display means 105, during a training phase of the apparatus operation to obtain a lasting display of the features present in a sample cough of a particular individual.
  • a human operator can manually program the decision logic 150 to recognize the coughs of the individual that supplied the sample and record the number of such coughs in a counter 155.
  • the decision logic 150 By programming the decision logic to the cough of an individual, it is possible to obtain high recognition accuracies. Also, the manual programming technique does not requireunduly expensive circuitry.
  • the translating means 20 is shown in further detail in FIG. 3 and is seen to include a preamplifier 21 the output of which provides parallel inputs into a bank 22 of fifteen contiguous bandpass filters, f through f
  • a preamplifier 21 the output of which provides parallel inputs into a bank 22 of fifteen contiguous bandpass filters, f through f
  • Each of the band-pass filters in the bank 22 receives the preamplified input signal and produces an output signal related to that portion of the input signal which lies in the range of frequencies passed by theparticular filter.
  • the filter center frequencies range from about 700 c.p.s. to about 6,500 c.p.s. with the lowest filter bandwidth being about 300 cycles.
  • the present embodiment is thus designed without particular sensitivity to frequencies below about 500 c.p.s. These lower frequencies can be quite useful in determining the presence of certain speech sounds, but it is found that coughs are not rich in the frequencies at the lower end of the speech spectrum, so'an economy of filters can be implemented herein.
  • each filter .in the bank 22 is individually coupled to a full wave rectifier and lowpass filter combination located in a rectifier/filter bank 23. After rectification and filtering, the outputs of the bank 23 essentially represent the energy levels of the input signal at about the center frequencies of each of the bandpass filters in the bank 22. Viewed in another way, the signals on lines 23a through 23p collectively represent the envelope of the energy vs. frequency of the received input signal taken over the frequency range of interest.
  • the fifteen channels of information on lines 23a through 23p are logarithmically compressed to produce the spectral component outputs on lines a through 20p of the translating means 20.
  • Logarithmic compression facilitates subsequent processing in two ways. First, it provides dynamic range compression that simplifies the engineering design requirements of feature signal generator 50. Secondly, by virtue of using logarithms, comparative ratios of the spectral component signals can be readily computed by subtraction. Ratios are desirable processing vehicles in that they are independent of changes in overall signal amplitudes. This property is particularly advantageous in a cough detection system where coughs of varying intensity are to be monitored.
  • a single log amplifier 25 is time shared to avoid the necessity of using fifteen identical amplifiers to achieve compression.
  • the outputs on lines 23a through 23p are effectively sampled by a multiplexer 24 and the sampled signals passed, one at a time, through the shared amplifier 25.
  • a demultiplexer 26 then reconstructs compressed spectral component signals on lines 20a through 20p from the processed sampled signals.
  • the sampling clock rate of the multiplexer and demultiplexer is above one megacycle and is safely higher than is necessary to retain signal bandwidths.
  • the spectral component signals on lines 20a through 20p are entered intothe feature signal generator 50 which senses the presence of properties of the spectral component signals that correspond to preselected properties or features of coughs.
  • this sensing of properties or feature extraction is achieved in part by deriving quantities known as positive and negative broad slope characteristics. These quantities give indication as to the polarity and magnitude of the slope of the input envelope (e. g., FIG. 1) when taken over specified relatively broad segments of frequency spectrum.
  • a broad positive slope at a particular frequency reference point is defined by the equation I n+1 n+2) n rr-1)] where K is a gain constant, the ES refer to the amplitudes of the processed spectral component signals on lines 20a through 20p, and n is a reference index that corresponds to the subscript index of the filters in the bank 22.
  • the broad positive slope at the reference frequency F (FIG. 1) is computed as A BPS has a non-zero value only when positive; i.e., when the first parenthetical term, (E E is greater than the second parenthetical term, (l5, E,, Otherwise the BPS is zero.
  • FIG. 4 there is shown a block diagram of the feature signal generator 50.
  • the spectral component signals on lines 200 through 20p are received from the translating means (FIG. 2) and coupled to a broad slope generator 51 that functions to develope BPS and BNS signals in the manner described above.
  • the bracket 52 encloses the signals available for further processing; i.e., the spectral component signals E, through E the broad positive slope signals BPS, through BPS,.,, and the broad negative slope signals BNS, through BNS
  • the interconnection of these signals to subsequent circuitry is not shown, it being understood that they are each available as inputs.
  • An operational amplifier 52 receives as excitory inputs the signals BNS,,, BNS BNS and BNS and receives as inhibitory inputs the signals BNS BNS ENS, and BNS
  • the amplifier 52 provides an analog output that is proportional to the amount by which the sum of the amplitudes of the signals at the excitory inputs exceed the sum of the amplitudes of the signals at the inhibitory inputs. Again, if the inhibitory input sum is greater than the excitory input sum, the amplifier output is zero.
  • the amplifier 52 thus produces an output only when selected broad negative slopes at the higher end of the considered frequency spectrum are collectively greater than selected broad negative slopes at the lower end of the considered frequency spectrum.
  • the property being detected i.e., a general negative slope or rollof at the higher end of the considered spectrum, has been found to be a common cough characteristic or feature.
  • Receiving the output of amplifier 52 is a NAND gate 52 which, absent an input of predetermined threshold from amplifier 52, produces a logical l output.
  • the gate 53 produces a logical O on its output line 53a, an occurrence that is designated as a Feature 5 signal.
  • the amplifier 54 and gate 55 similarly act to produce a Feature 6 signal on line 550 when another selected portion of the high frequency rolloff (BNS through BNS,,) exceeds any existent rolloff at another portion of low frequency spectrum (BNS through BNS).
  • the amplifier 56 compares spectral component relative energies, producing an output proportional to the amount by which the sum of E E and E exceeds the sum of E,,, E and 15, When the output of amplifier 56 is above a predetermined threshold level, the gate 57 produces a logical 0 that is designated as a Feature 7 signal on line 57a.
  • Feature 7 is, again, a measure of a determined cough characteristic, i.e., a particular energy ratio at the lower end of the considered frequency spectrum.
  • the remainder of the amplifiers 58, 60, 62, 64 70, and corresponding gates 59, 61, 63, 71, receive spectral component signals or broad slope signals and can produce signals on lines 59a, 61a, 63a 71a, that are indicative of Features 8 through 14.
  • Features 10 and 14 detect the presence of particular broad slope characteristics of interest.
  • the feature generator 50 generates a group of features signals that are subsequently used in a special manner in the present embodiment. These special features are designated herein as Feature 1 through Feature 4.
  • the Features 2 through 4 detect the same characteristics detected by the Features 5 through 10, but in a less restrictive manner.
  • the outputs of NAND gates 53 and 55 are coupled to the inputs of an AND gate 72.
  • the output of gate 72 is also a logical 0.
  • a l0gical 0 signal on output line 72a designated as Feature 2 will occur when either Feature 5 or Feature 6 is present.
  • Feature 1 is present when the sum of spectral component signals B, through E exceeds a predetermined threshold. (Recalling that we are dealing with logs, a product of energies is actually represented.) This feature detects the broad spread of substantial energy that is characteristic of coughs.
  • the inputs E through E,; are received by operational amplifier 75 which produces an output when the sum of the inputs exceeds a predetermined threshold level. This output is fed to NAND gate 76 which produces a logical O on its output line 76a, designated a Feature 1 signal, when the threshold condition is met.
  • FIG. 5 there are shown block diagrams of the storage means 100, the display means 105, and the occurrence decision logic 150.
  • the lines carrying the signals that represent Feature 1 through Feature 4 are received by flip-flops 111 through 114 in storage means 100.
  • These flip-flops which initially have a low output level by virtue of a manual reset signal on a line 106, are of the type that are set to a high output by an input transistion from a logic l to a logical 0.
  • the signals that represent Feature 5 through Feature 14 are coupled, via OR gates 125 through 134, to another group of the same type flip-flops, 115 through 124.
  • the other input to each of the OR gates through 134 is an enable signal received over a common line 136.
  • the signal on line 136 is normally at a logical 1 level so that the outputs of the OR gates 125 through 134 are maintained at l and the flipflops 115 through 124 are inhibited from being set to a high output state.
  • the outputs of the flip-flops 111 through 124 are coupled to lamp drivers D, through D,., in display means 105 which, in turn, activate the respective lamps L, through L
  • the lamps L numbered by the features to which they correspond, can be conveniently displayed on the faceplate of a'console 300 that houses produces an output only after receiving an input of a specified minimum duration.
  • the lines from switches S, through S, are also coupled as inputs. to another NOR gate 165, the output of which is fed to an enable generator that preferably comprises a one-shot multivibrator.
  • the output of the enable generator 170 is infaceplate of console 300 (FIG. 6), adjacent their correspondingly numbered lamps.
  • the manual reset line 106 is coupled to a logic level 1 supply (not shown) through a pushbutton switch on the console
  • the operation of the system typically commences by programming or training the apparatus to the cough of a particular individual. First, an operator puts all switches 8, through S in the closed (down) position. The reset button on the console is then pressed and the resultant signal on line 106 (FIG. 5) resets all the flipflops 111 through 124 so that the lamps L through L are all off. At this point, assuming for the moment that ambient room noise is zero, the signals which represent Feature 1 through Feature 14 (FIGS. 4 and 5) are all at a logic l level and, consequently, the outputs of the NOR gate 155 and 165 are at a logic 0. Thus, the output of the enable generator 170 is at so that the output of inverter 175 is at l, effectively disabling flip-flops 115 through 124 by action of line 136.
  • the input cough contains characteristics that activate Feature 1 through Feature 4 (which is likely) in addition to certain other Features, say 5, 7, 9, 10, 11', 13 and 14. If this occurs, the signals corresponding to these features change to a logic 0 and, as one consequence, the four inputs to NOR gate 165 are at 0" so the output of NOR gate 165 goes to l
  • the enable generator producing a l output for a predetermined duration (preferably about I second) so that a 0 enable signal appears on line 136 for about I second.
  • the operator observes which lamps are not lit and opens the switches S that correspond to the unlit lamps.
  • the apparatus is then programmed for the cough of the individual that supplied the sample cough.
  • NOR gate 155 lasts for longer than the predetermined duration of sensor 160 (i.e., if all the inputs 1 through 5, 7, 9, 10, 11, 13 and 14 last for this predetermined duration), the sensor 160 will produce an output that trips the counter 190 (FIG. 2).
  • Variations in the programming procedure can be implemented to further reduce the probability of extraneous triggering of the counter or to increase the sensitivity of the apparatus.
  • a number of training coughs can be received during the programming phase (normally resetting the lamps after each one) and the swtiches S opened for any feature that does not occur during every training cough.
  • Another technique for modifying the sensitivity of the apparatus to a desired level is adjustment of the duration sensor critical time. This is conveniently accomplished during the programming phase by a trial and error technique of determining the maximum setting at which a patients cough trips the counter.
  • NOR gate 155 The output of NOR gate 155 is received by an adjustable one-short multivibrator 161 that is preferably adjustable to remain in its unstable state for between about 50 ms. to 250 ms. from its last triggering.
  • the one-shot 161 is of the type that has an output that is normally high (a logical l) until triggered by the edge of a positivegoing signal whereupon its output goes to a logical 0 level.
  • the output of one-shot 161 is received by AND gate 162.
  • Theoutput of NOR gate 155 is also coupled, via a short delay means 163, to AND gate 162.
  • the delay means 163 need only provide a delay which is longer than the propagation time of the one-shot 161 and may have, for example, an intrinsic delay of about nanoseconds.
  • Another one-shot 164 which is of the same type as one-shot 161 but has a longer and fixed unstable state of about lsecond, receives the output of AND gate 162. The output of one-shot 164 is fed back as an input (designated 164a) to AND gate 162 and is also received by the counter 190 (FIG. 2) which is of the type that is triggered by the edge of a negative going signal.
  • duration sensor circuit 160 The operation of duration sensor circuit 160 is as follows: Assume that the adjustable one-shot 161 is set at a 250 ms. duration. When the output of NOR gate (designated 155a) is 0 the AND gate 162 is disabled by its center input (designated 16311). When 155a goes to l, the output of one-shot 161 (designated 161a) goes to 0 and remains there for 250 ms. The delay 163 prevents 163a from being at 1 until after 161a has gone to 0 so that both these signals cant be at l simultaneously'at the onset of a l on line 155a. After 250 ms.
  • the one-shot 161 will return to its stable output of l If 155a has remained at 1 during this 250 ms., the three inputs to gate 162 will be at 1 simultaneously and the resultant output of gate 162 will trigger one-shot 164 to a state which, in turn, trips the counter. The fed-back 0 on line 164a then prevents the one-shot 164a from triggering again for at least 1 second so that no more than one cough per second can be counted. Also, it should be noted that, in order to trigger the one-shot 164, the input 155a must be at 1 continually for the specified time of 250 ms. If 155a returns to O and then goes back to l during a given 250 ms. interval, the one-shot 161 will start a new 250 ms. counting period from the latest excursion to 1.
  • the duration sensor 160 (FIG. 5) can be adjusted to the maximum setting at which a patients cough is found to trip the counter. By so doing, the probability of the counter being triggered by extraneous noise is reduced even further.
  • the preferred procedure is to start at a setting of 250 ms. and, after programming, observe whether sample coughs trip the counter. If not, successively shorter settings can be tried until appropriate results are obtained.
  • the duration sensor 160 may be designed to produce an output for a less restrictive condition that requires an input only at the beginning and end of a specified period rather than continuously during such period.
  • said feature signal generating means includes means for sensing a predetermined minimum energy product in the spectral component signals.
  • Apparatus as defined by claim 1 further comprising duration sensing means which enables occurrence indications only when the predetermined combination of feature signals is continuously present for a predetermined duration.
  • said occurrence decision means includes a plurality of manually operable switches and a gate, each of said switches coupling a feature signal to said gate, the presence of a feature signal on all gate inputs that are coupled through closed switches causing activation of said gate.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

An apparatus that receives ambient sounds and recognizes the occurrence of coughs to the exclusion of other sounds. Means are provided for translating the received sounds into a plurality of spectral component signals. A feature signal generating means is provided for sensing the presence of properties of the spectral component signals that are characteristic of coughs and for generating a feature signal for each property found to be present. An occurrence decision means, responsive to the feature signals, generates an occurrence indication upon receiving a predetermined combination of feature signals. In a preferred embodiment of the invention, the occurrence decision means is programmable to the cough of a particular individual.

Description

United States Patent 1191 Herscher et al.
1111 3,821,472 [4 June 28, 1974 [54] COUGH MONITORING APPARATUS 3,588,363 6/197] Herscher 179/1 SA [75] Inventors: Marvin B. Herscher, Camden; igggggg Thomas B. Martin, Burlington, both of Primary Examiner-Kathleen H. Claffy [73] Assignee: Threshold Technology Inc., Assistant Examiner.lon Bradford Leaheey Cinnaminson, NJ.
[22] Filed: Feb. 25, 1972 ABSTRACT 2 App] 229,290 An apparatus that receives ambient sounds and recognizes the occurrence of coughs to the exclusion of other sounds. Means are provided for translating the [52] US. Cl 179/1 SA received sounds into a plurality of Spectra] Component [51] Int. Cl. G] l/04 signals A feature signal generating meanslis provided [58] Field of Search 179/ l 1 1 for sensing the presence of properties of the spectral Vci 128/2 R121 component signals that are characteristic of coughs and for generating a feature signal for each property [56] References C'ted found to be present. An occurrence decision means, UNITED STATES PATENTS responsive to the feature signals, generates an occur- 3,198,884 8/1965 Dersch 179/1 SA rence indication p receiving a predetermined cOrn- 3,204,030 8/1965 Olson 179/1 SB bination of feature signals. In a preferred embodiment 3,287,649 11/1966 Rosenblatt l79/l SA of the invention, the occurrence deci i n.means is 3,368,039 2/1968 Clapper 179/1 SA ble to the cough of a particular individual. 3,445,594 5/1969 Kusch l79/l SA 3,466,394 9/1969 French 179/1 SB 5 Claims, 7 Drawing Figures DISPLAY 20 o 20 so I} sromcz 5 5 20b I 20c E 20d t 3011 l FEA'iURE TRANSLATING I MEANS i GEiJ Si A T OR PROGRAMMING l I I l 20 l g OCCURRENCE I LOGIC SPECTRAL FEATURE COMPONENT SIGNALS I SIGNALS l PATENTEDJUHZB I974 l0 ll IZ sum 1 or s ENERGY FREQUENCY Fig. I.
PATENTEDJUHZB I974 SHEET 5 (IF 5 MANUAL RESET 4 l 2 3 4 l L L L L L 1 3 3 E 4 l 2 3 4 5 I D D D D D T A R PE PR R R R W MV MV M AMA A H llll. MV Am A A R R LR LR R LO O O D D V 4 w w m w 2 I I rh F. F F. F. L L J A 6 O 4 s 5 llll Q I Z 3 4 5 M E E E E E E R R R R R R u U m U m u T T T M A A A A A E E E E E E F F F F F F ENABLE GEN TO COUNTER OR SENSOR DURATION RECORDER FEATURE FEATURE 2 FEATURE 4 FEATURE l4 ADJUST Fig. 5.
1 COUGH MONITORING APPARATUS BACKGROUND OF THE INVENTION is assigned the monitoring task there are obvious disadvantages of inconvenience, cost and unreliability.
Equipments that attempt to automatically monitor humancough rates have been previously developed, but are found to suffer various deficiencies. As an example, one prior art device includes a pneumatic band that is fitted around a patients chest and connected to a recorder through pneumatic tubes. Problems of patient discomfort limit this devices usefulness. Another scheme provides a microphone carried on a patients person or positioned nearby. Sounds received by the microphone are transmitted as electrical signals to a receiver which triggers a counter whenever the power of the received signals exceeds a predetermined level. The
drawback of this system is triggering by sounds other than the coughs of the patient being monitored. Conversation and other room noises can cause such extraneous triggering with a resultant incorrect cough count.
In view of the disadvantages of prior art systems, it is an object of the present invention to provide an improved apparatus which detects coughs with reliability and without patient discomfort.
SUMMARY OF THE INVENTION The present invention is directed to an apparatus which receives ambient sounds and reliably recognizes the occurrence of specified non-verbal sounds, e.g., coughs, to the exclusion of other sounds. Means are provided for translating the received sounds into a plurality of spectral component signals. A feature signal generating means is provided for sensing the presence of properties of the spectral component signals that are characteristic of coughs and for generating a feature signal for each property found to be present. An occurrence decision means, responsive to the feature signals, generates an occurrence indication upon receiving a predetermined combination of feature signals.
In a preferred embodiment of the invention, the occurrence decision means is programmable for different specified sounds, for example, the cough of a particular individual.
Further features and advantages of the invention will become more readily apparent from the following detailed description when taken in conjunction with the accompanying drawings.
DESCRIPTION OF THE DRAWINGS FIG. 1 is a graphical representation of the general shape of the energy vs. frequency spectrum for a typical human cough; I
FIG. 2 is a block diagram of an embodiment of a cough monitoring apparatus in accordance with the invention;
FIG. 3 is a block diagram of the translating means portion of the invented apparatus;
FIG. 4 is a block diagram representation that illustrates the functioning of the feature signal generator utilized in the present invention;
FIG. 5 is a block diagram of the storage means and occurrence decision logic means utilized in the present invention.
FIG. 6 is a side perspective view of a console utilized to house an embodiment of the present invention.
FIG. 7 is a block diagram of the duration sensor utilized in the occurrence decision logic means of FIG. 5.
DESCRIPTION OF THE PREFERRED EMBODIMENTS Referring to FIG. 1, there is shown an energy vs. frequency plot that illustrates the typical instantaneous shape of a human cough. The arrows at F, through F represent the energy levels at selected frequencies of a cough and the dashed line represents the spectrum envelope.
It has been found that coughs have certain combinations of distinguishing characteristics that are usually not present in speech or ambient room noise. Generally, a cough is noiselike burst of energy of substantial duration characterized by a sharp onset and having a single broad peak in its energy vs. frequency spectrum as shown in FIG.'1. The frequency components that contain most of the cough energy are in the range about 500-7500 c.p.s., and the broad peak is typically in the range about l,0003,000 c.p.s. Many speech sounds, on the other hand, have two or more distinct peaks or formants in their energy vs. frequency. Also, it has been empirically determined that certain slope and energy relationships exist between the different spectral components of most coughs. By determining which of a judiciously selected set of relationships or features is present in the cough of a particular individual, and then programming the invented apparatus to sensethe presence of the determined combination, it is possible to. achieve recognition accuracies that were heretofore unattainable.
Referring now to FIG. 2, there is shown a block diagram of an embodiment of a cough recognition system in accordance with the invention. A tranducer 30, typically a gradient microphone, receives ambient input sounds and produces time-varying electrical signals representative of the received sounds. (Hereinafter, the phrases input sounds or ambient input sounds are generically defined as including sound-representative signals, for example, signals from a tape recorder that were taken at a patients bedside for later processing.) The sound representative signals are coupled over' a line 30a to a translating means 20 that includes, inter alia, a bank of frequency-selective devices such as bandpass filters. The output of the translating means is a plurality of signals on lines 20a through 20p. These output signals represent spectral components of the received sounds and can be pictured, for example, as reflecting the energy levels at the selected frequencies shown in FIG. 1. v
A feature signal generating means 50 receives the spectral component signals 20a through 20p and senses the presence of properties of the spectral component signals that correspond to preselected properties or features of coughs. The manner in which this is achieved is described in further detail hereinbelow. It suffices for the present to understand that each of the output lines corresponds to one of the preselected feameans 100 and occurrence decision logic means 150.
Storage means 100 is utilized, in conjunction with display means 105, during a training phase of the apparatus operation to obtain a lasting display of the features present in a sample cough of a particular individual. Assisted by the display, a human operator can manually program the decision logic 150 to recognize the coughs of the individual that supplied the sample and record the number of such coughs in a counter 155. By programming the decision logic to the cough of an individual, it is possible to obtain high recognition accuracies. Also, the manual programming technique does not requireunduly expensive circuitry.
The translating means 20 is shown in further detail in FIG. 3 and is seen to include a preamplifier 21 the output of which provides parallel inputs into a bank 22 of fifteen contiguous bandpass filters, f through f Each of the band-pass filters in the bank 22 receives the preamplified input signal and produces an output signal related to that portion of the input signal which lies in the range of frequencies passed by theparticular filter. In the present embodiment, the filter center frequencies range from about 700 c.p.s. to about 6,500 c.p.s. with the lowest filter bandwidth being about 300 cycles. The present embodiment is thus designed without particular sensitivity to frequencies below about 500 c.p.s. These lower frequencies can be quite useful in determining the presence of certain speech sounds, but it is found that coughs are not rich in the frequencies at the lower end of the speech spectrum, so'an economy of filters can be implemented herein.
The output of each filter .in the bank 22 is individually coupled to a full wave rectifier and lowpass filter combination located in a rectifier/filter bank 23. After rectification and filtering, the outputs of the bank 23 essentially represent the energy levels of the input signal at about the center frequencies of each of the bandpass filters in the bank 22. Viewed in another way, the signals on lines 23a through 23p collectively represent the envelope of the energy vs. frequency of the received input signal taken over the frequency range of interest. I
The fifteen channels of information on lines 23a through 23p are logarithmically compressed to produce the spectral component outputs on lines a through 20p of the translating means 20. Logarithmic compression facilitates subsequent processing in two ways. First, it provides dynamic range compression that simplifies the engineering design requirements of feature signal generator 50. Secondly, by virtue of using logarithms, comparative ratios of the spectral component signals can be readily computed by subtraction. Ratios are desirable processing vehicles in that they are independent of changes in overall signal amplitudes. This property is particularly advantageous in a cough detection system where coughs of varying intensity are to be monitored.
In the diagram of FIG. 3, a single log amplifier 25 is time shared to avoid the necessity of using fifteen identical amplifiers to achieve compression. The outputs on lines 23a through 23p are effectively sampled by a multiplexer 24 and the sampled signals passed, one at a time, through the shared amplifier 25. A demultiplexer 26 then reconstructs compressed spectral component signals on lines 20a through 20p from the processed sampled signals. The sampling clock rate of the multiplexer and demultiplexer is above one megacycle and is safely higher than is necessary to retain signal bandwidths. This technique of sharing a single logarithmicamplifier is known in the art and is disclosed, for example, in US. Pat. No. 3,588,363 of M. Herscher and T. Martin entitled Word Recognition System for Voice Controller.
Referring again, momentarily, to FIG. 2, it will be recalled that the spectral component signals on lines 20a through 20p are entered intothe feature signal generator 50 which senses the presence of properties of the spectral component signals that correspond to preselected properties or features of coughs. In the present embodiment, this sensing of properties or feature extraction is achieved in part by deriving quantities known as positive and negative broad slope characteristics. These quantities give indication as to the polarity and magnitude of the slope of the input envelope (e. g., FIG. 1) when taken over specified relatively broad segments of frequency spectrum.
The manner in which broad positive slopes and braod negative slopes are obtained is described in detail in the above-referenced US. Patent of Herscher and Martin. A broad positive slope at a particular frequency reference point is defined by the equation I n+1 n+2) n rr-1)] where K is a gain constant, the ES refer to the amplitudes of the processed spectral component signals on lines 20a through 20p, and n is a reference index that corresponds to the subscript index of the filters in the bank 22. For example, the broad positive slope at the reference frequency F (FIG. 1) is computed as A BPS has a non-zero value only when positive; i.e., when the first parenthetical term, (E E is greater than the second parenthetical term, (l5, E,, Otherwise the BPS is zero.
The broad negative slopes are defined by BN8. KuE. E.-.) w... E...)]
and similarly have non-zero values only when positive. The physical implementation of the equations for the broad positive slopes and broad negative slopes is accomplished by using operational amplifiers. These units, when fitted with appropriate peripheral circuit components, provide analog output signals which are proportional to the difference between the sum of the amplitudes of signals at excitory input terminals and the sum of the amplitudes of signals at inhibitory terminals. With 15 spectral component signals, E through E there are 14 possible broad positive slopes from BPS, to BPS and 14 corresponding broad negative slopes from BNS, to ENS (Calculations of BPS and BNS reduce to zero and I((E E respectively, and therefore provide no meaningful slope information.)
Referring to FIG. 4, there is shown a block diagram of the feature signal generator 50. The spectral component signals on lines 200 through 20p are received from the translating means (FIG. 2) and coupled to a broad slope generator 51 that functions to develope BPS and BNS signals in the manner described above. The bracket 52 encloses the signals available for further processing; i.e., the spectral component signals E, through E the broad positive slope signals BPS, through BPS,.,, and the broad negative slope signals BNS, through BNS For clarity of illustration, the interconnection of these signals to subsequent circuitry is not shown, it being understood that they are each available as inputs.
An operational amplifier 52 receives as excitory inputs the signals BNS,,, BNS BNS and BNS and receives as inhibitory inputs the signals BNS BNS ENS, and BNS In the manner previously described, the amplifier 52 provides an analog output that is proportional to the amount by which the sum of the amplitudes of the signals at the excitory inputs exceed the sum of the amplitudes of the signals at the inhibitory inputs. Again, if the inhibitory input sum is greater than the excitory input sum, the amplifier output is zero. The amplifier 52 thus produces an output only when selected broad negative slopes at the higher end of the considered frequency spectrum are collectively greater than selected broad negative slopes at the lower end of the considered frequency spectrum. The property being detected, i.e., a general negative slope or rollof at the higher end of the considered spectrum, has been found to be a common cough characteristic or feature. Receiving the output of amplifier 52 is a NAND gate 52 which, absent an input of predetermined threshold from amplifier 52, produces a logical l output. When the output of amplifier 52 exceeds a predetermined threshold level, the gate 53 produces a logical O on its output line 53a, an occurrence that is designated as a Feature 5 signal.
The amplifier 54 and gate 55 similarly act to produce a Feature 6 signal on line 550 when another selected portion of the high frequency rolloff (BNS through BNS,,) exceeds any existent rolloff at another portion of low frequency spectrum (BNS through BNS The amplifier 56 compares spectral component relative energies, producing an output proportional to the amount by which the sum of E E and E exceeds the sum of E,,, E and 15, When the output of amplifier 56 is above a predetermined threshold level, the gate 57 produces a logical 0 that is designated as a Feature 7 signal on line 57a. Feature 7 is, again, a measure of a determined cough characteristic, i.e., a particular energy ratio at the lower end of the considered frequency spectrum.
The remainder of the amplifiers 58, 60, 62, 64 70, and corresponding gates 59, 61, 63, 71, receive spectral component signals or broad slope signals and can produce signals on lines 59a, 61a, 63a 71a, that are indicative of Features 8 through 14. Features 8, 9, and 11, for example, detect the presence of the relatively broad peak of energy at about the middle of the considered spectrum, a characteristic found present in most coughs. Features 10 and 14 detect the presence of particular broad slope characteristics of interest.
In addition to Features 5 through 14, the feature generator 50 generates a group of features signals that are subsequently used in a special manner in the present embodiment. These special features are designated herein as Feature 1 through Feature 4. The Features 2 through 4 detect the same characteristics detected by the Features 5 through 10, but in a less restrictive manner. For example, the outputs of NAND gates 53 and 55 are coupled to the inputs of an AND gate 72. When either of the signals on lines 53a or 55a is a logical O, the output of gate 72 is also a logical 0. Thus, a l0gical 0 signal on output line 72a, designated as Feature 2, will occur when either Feature 5 or Feature 6 is present. The same relationship exists as between Feature 3 (the output of gate 73 on line 73a) with Features 7 and 8, and as between Feature 4 (the output of gate 74 on line 74a) with Features 9 and 10. In other words, F eature 3 will be present whenever either Feature 7 or Feature 8 is present, and Feature 4 will be present whenever either Feature 9 or Feature 10 is present.
The remaining special feature is Feature 1, which is present when the sum of spectral component signals B, through E exceeds a predetermined threshold. (Recalling that we are dealing with logs, a product of energies is actually represented.) This feature detects the broad spread of substantial energy that is characteristic of coughs. The inputs E through E,;, are received by operational amplifier 75 which produces an output when the sum of the inputs exceeds a predetermined threshold level. This output is fed to NAND gate 76 which produces a logical O on its output line 76a, designated a Feature 1 signal, when the threshold condition is met.
Referring to FIG. 5, there are shown block diagrams of the storage means 100, the display means 105, and the occurrence decision logic 150. The lines carrying the signals that represent Feature 1 through Feature 4 are received by flip-flops 111 through 114 in storage means 100. These flip-flops, which initially have a low output level by virtue of a manual reset signal on a line 106, are of the type that are set to a high output by an input transistion from a logic l to a logical 0. The signals that represent Feature 5 through Feature 14 are coupled, via OR gates 125 through 134, to another group of the same type flip-flops, 115 through 124. The other input to each of the OR gates through 134 is an enable signal received over a common line 136. The signal on line 136 is normally at a logical 1 level so that the outputs of the OR gates 125 through 134 are maintained at l and the flipflops 115 through 124 are inhibited from being set to a high output state.
The outputs of the flip-flops 111 through 124 are coupled to lamp drivers D, through D,., in display means 105 which, in turn, activate the respective lamps L, through L The lamps L, numbered by the features to which they correspond, can be conveniently displayed on the faceplate of a'console 300 that houses produces an output only after receiving an input of a specified minimum duration. The lines from switches S, through S, are also coupled as inputs. to another NOR gate 165, the output of which is fed to an enable generator that preferably comprises a one-shot multivibrator. The output of the enable generator 170 is infaceplate of console 300 (FIG. 6), adjacent their correspondingly numbered lamps. Also, the manual reset line 106 is coupled to a logic level 1 supply (not shown) through a pushbutton switch on the console The operation of the system typically commences by programming or training the apparatus to the cough of a particular individual. First, an operator puts all switches 8, through S in the closed (down) position. The reset button on the console is then pressed and the resultant signal on line 106 (FIG. 5) resets all the flipflops 111 through 124 so that the lamps L through L are all off. At this point, assuming for the moment that ambient room noise is zero, the signals which represent Feature 1 through Feature 14 (FIGS. 4 and 5) are all at a logic l level and, consequently, the outputs of the NOR gate 155 and 165 are at a logic 0. Thus, the output of the enable generator 170 is at so that the output of inverter 175 is at l, effectively disabling flip-flops 115 through 124 by action of line 136.
A sample cough of a patient, taken either live or from a recording, is next entered into the translating means, for example, via transducer 30. Assume for the moment that the input cough contains characteristics that activate Feature 1 through Feature 4 (which is likely) in addition to certain other Features, say 5, 7, 9, 10, 11', 13 and 14. If this occurs, the signals corresponding to these features change to a logic 0 and, as one consequence, the four inputs to NOR gate 165 are at 0" so the output of NOR gate 165 goes to l This results in the enable generator producing a l output for a predetermined duration (preferably about I second) so that a 0 enable signal appears on line 136 for about I second. The OR gates 125 through 134 are thus enabled to pass the 0 levels of the activated features among Feature 5 through 14 to set the corresponding flip-flops among flip-flops 115 through 124.. Also, the flip-flops 111 through 114 are set by the transitions to O of the signals representative of Feature 1 through Feature 4. As a result, all lamps corresponding to the occurring features are lit and stay lit. (It should be noted at this point that the sample cough does not activate NOR gate 155 since, with all switches closed, some of the inputs to the gate are at a logic l level; i.e., the inputs corresponding to Features 6, 8, and 12 for the example given.)
The operator observes which lamps are not lit and opens the switches S that correspond to the unlit lamps. The apparatus is then programmed for the cough of the individual that supplied the sample cough. To illustrate, if, in the example given, the switches S S and S are opened during programming, a subsequent cough by the patient that contains the Features 1 through 5, 7, 9, 11, 13 and 14 will cause a 1 output from NOR gate 155 since this gate will now have no inputs at the logic l level. If the output of NOR gate 155 lasts for longer than the predetermined duration of sensor 160 (i.e., if all the inputs 1 through 5, 7, 9, 10, 11, 13 and 14 last for this predetermined duration), the sensor 160 will produce an output that trips the counter 190 (FIG. 2).
From the above description it is seen that the absence of an enable signal on line 136 prevents the lamps L through L from being individually lit by extraneous ambient noise that occurs during the programming phase of operation. It is only when Features 1 through 4 occur simultaneously (and this is highly unlikely except during a cough) that the lamps L through L, can be lit. Thus, the probability of an improper programming is minimized.
There is the possibility that the coughs of certain individuals will not contain all of the Features 1 through 4. If this occurs, when the training sample cough is received only some of the lamps L through L will be lit, say, L, through L In such event, the operator can open switch S effectively introducing a logic 0 input on the Feature 4 input line to NOR gate 165. Now, the enable signal on line 136 can be triggered by the occurrence of Features 1 through 3, so the operator can press the manual reset button and receive another training cough to accomplish a proper programming.
Variations in the programming procedure can be implemented to further reduce the probability of extraneous triggering of the counter or to increase the sensitivity of the apparatus. For example, a number of training coughs can be received during the programming phase (normally resetting the lamps after each one) and the swtiches S opened for any feature that does not occur during every training cough. Another technique for modifying the sensitivity of the apparatus to a desired level is adjustment of the duration sensor critical time. This is conveniently accomplished during the programming phase by a trial and error technique of determining the maximum setting at which a patients cough trips the counter.
Referring to FIG. 7, there is shown a simplified block diagram of the duration sensor 160. The output of NOR gate 155 is received by an adjustable one-short multivibrator 161 that is preferably adjustable to remain in its unstable state for between about 50 ms. to 250 ms. from its last triggering. The one-shot 161 is of the type that has an output that is normally high (a logical l) until triggered by the edge of a positivegoing signal whereupon its output goes to a logical 0 level. The output of one-shot 161 is received by AND gate 162. Theoutput of NOR gate 155 is also coupled, via a short delay means 163, to AND gate 162. The delay means 163 need only provide a delay which is longer than the propagation time of the one-shot 161 and may have, for example, an intrinsic delay of about nanoseconds. Another one-shot 164, which is of the same type as one-shot 161 but has a longer and fixed unstable state of about lsecond, receives the output of AND gate 162. The output of one-shot 164 is fed back as an input (designated 164a) to AND gate 162 and is also received by the counter 190 (FIG. 2) which is of the type that is triggered by the edge of a negative going signal.
The operation of duration sensor circuit 160 is as follows: Assume that the adjustable one-shot 161 is set at a 250 ms. duration. When the output of NOR gate (designated 155a) is 0 the AND gate 162 is disabled by its center input (designated 16311). When 155a goes to l, the output of one-shot 161 (designated 161a) goes to 0 and remains there for 250 ms. The delay 163 prevents 163a from being at 1 until after 161a has gone to 0 so that both these signals cant be at l simultaneously'at the onset of a l on line 155a. After 250 ms. have elapsed from the time that 1550 went to l, the one-shot 161 will return to its stable output of l If 155a has remained at 1 during this 250 ms., the three inputs to gate 162 will be at 1 simultaneously and the resultant output of gate 162 will trigger one-shot 164 to a state which, in turn, trips the counter. The fed-back 0 on line 164a then prevents the one-shot 164a from triggering again for at least 1 second so that no more than one cough per second can be counted. Also, it should be noted that, in order to trigger the one-shot 164, the input 155a must be at 1 continually for the specified time of 250 ms. If 155a returns to O and then goes back to l during a given 250 ms. interval, the one-shot 161 will start a new 250 ms. counting period from the latest excursion to 1.
As above-stated, the duration sensor 160 (FIG. 5) can be adjusted to the maximum setting at which a patients cough is found to trip the counter. By so doing, the probability of the counter being triggered by extraneous noise is reduced even further. The preferred procedure is to start at a setting of 250 ms. and, after programming, observe whether sample coughs trip the counter. If not, successively shorter settings can be tried until appropriate results are obtained.
While the invention has been described above with reference to a particular embodiment, it will be appreciated that variations are available within the spirit of the invention. As an example, the duration sensor 160 may be designed to produce an output for a less restrictive condition that requires an input only at the beginning and end of a specified period rather than continuously during such period.
We claim:
1. Apparatus for receiving ambient input sounds and recognizing the occurrence of coughs to the exclusion of other sounds, comprising:
a. means for translating said input sounds into a plurality of spectral component signals;
b. feature signal generating means for sensing the properties of a single broad peak of energy in the spectral component signals and for generating a feature signal for each such property found to be present; and p c. occurrence decision means responsive to said feature signals for generating occurrence indications upon receiving a predetermined combination of said feature signals, irrespective of the order of arrival of said feature signals.
2. Apparatus as defined in claim 1 wherein said feature signal generating means includes means for sensing a predetermined minimum energy product in the spectral component signals.
3. Apparatus as defined by claim 1 further comprising duration sensing means which enables occurrence indications only when the predetermined combination of feature signals is continuously present for a predetermined duration.
4. Apparatus as defined in claim 1 wherein said occurrence decision means includes a plurality of manually operable switches and a gate, each of said switches coupling a feature signal to said gate, the presence of a feature signal on all gate inputs that are coupled through closed switches causing activation of said gate.
mined duration.

Claims (5)

1. Apparatus for receiving ambient input sounds and recognizing the occurrence of coughs to the exclusion of other sounds, comprising: a. means for translating said input sounds into a plurality of spectral component signals; b. feature signal generating means for sensing the properties of a single broad peak of energy in the spectral component signals and for generating a feature signal for each such property found to be present; and c. occurrence decision means responsive to said feature signals for generating occurrence indications upon receiving a predetermined combination of said feature signals, irrespective of the order of arrival of said feature signals.
2. Apparatus as defined in claim 1 wherein said feature signal generating means includes means for sensing a predetermined minimum energy product in the spectral component signals.
3. Apparatus as defined by claim 1 further comprising duration sensing means which enables occurrence indications only when the predetermined combination of feature signals is continuously present for a predetermined duration.
4. Apparatus as defined in claim 1 wherein said occurrence decision means includes a plurality of manually operable switches and a gate, each of said switches coupling a feature signal to said gate, the presence of a feature signal on all gate inputs that are coupled through closed switches causing activation of said gate.
5. Apparatus as defined by claim 4 wherein said occurrence decision means further includes duration sensing means which allow occurrence indications when said gate is continuously activated for a predetermined duration.
US00229290A 1972-02-25 1972-02-25 Cough monitoring apparatus Expired - Lifetime US3821472A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US00229290A US3821472A (en) 1972-02-25 1972-02-25 Cough monitoring apparatus

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US00229290A US3821472A (en) 1972-02-25 1972-02-25 Cough monitoring apparatus

Publications (1)

Publication Number Publication Date
US3821472A true US3821472A (en) 1974-06-28

Family

ID=22860579

Family Applications (1)

Application Number Title Priority Date Filing Date
US00229290A Expired - Lifetime US3821472A (en) 1972-02-25 1972-02-25 Cough monitoring apparatus

Country Status (1)

Country Link
US (1) US3821472A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5201028A (en) * 1990-09-21 1993-04-06 Theis Peter F System for distinguishing or counting spoken itemized expressions
US5315688A (en) * 1990-09-21 1994-05-24 Theis Peter F System for recognizing or counting spoken itemized expressions
US20020059064A1 (en) * 2000-11-10 2002-05-16 Hajime Tabata Speech communication apparatus
US20060277037A1 (en) * 2005-06-03 2006-12-07 Woodcock Ashley A Method for generating output data
EP1879501A2 (en) * 2005-04-29 2008-01-23 Oren Gavriely Cough detector
CN101894551A (en) * 2010-07-02 2010-11-24 华南理工大学 Method and device for automatically identifying cough
CN109498228A (en) * 2018-11-06 2019-03-22 林枫 Lung recovery therapeutic equipment based on cough sound feedback

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5201028A (en) * 1990-09-21 1993-04-06 Theis Peter F System for distinguishing or counting spoken itemized expressions
US5315688A (en) * 1990-09-21 1994-05-24 Theis Peter F System for recognizing or counting spoken itemized expressions
US5577163A (en) * 1990-09-21 1996-11-19 Theis; Peter F. System for recognizing or counting spoken itemized expressions
US20020059064A1 (en) * 2000-11-10 2002-05-16 Hajime Tabata Speech communication apparatus
US7107209B2 (en) * 2000-11-10 2006-09-12 Honda Giken Kogyo Kabushiki Kaisha Speech communication apparatus with gain control for clear communication
EP1879501A2 (en) * 2005-04-29 2008-01-23 Oren Gavriely Cough detector
EP1879501A4 (en) * 2005-04-29 2009-05-20 Oren Gavriely Cough detector
US20060277037A1 (en) * 2005-06-03 2006-12-07 Woodcock Ashley A Method for generating output data
US7761302B2 (en) 2005-06-03 2010-07-20 South Manchester University Hospitals Nhs Trust Method for generating output data
CN101894551A (en) * 2010-07-02 2010-11-24 华南理工大学 Method and device for automatically identifying cough
CN109498228A (en) * 2018-11-06 2019-03-22 林枫 Lung recovery therapeutic equipment based on cough sound feedback

Similar Documents

Publication Publication Date Title
US4069393A (en) Word recognition apparatus and method
US4032710A (en) Word boundary detector for speech recognition equipment
CA2773294C (en) Sound detection and localization system
EP0060086B1 (en) Speech recognition system and method
US6895098B2 (en) Method for operating a hearing device, and hearing device
Dalland Hearing sensitivity in bats
EP0077194B1 (en) Speech recognition system
US2761897A (en) Electronic device for automatically discriminating between speech and music forms
US4770189A (en) Real time multitask electronic stethoscopy system
US4668941A (en) Method and apparatus for discriminating sounds due to the breakage or glass
US5608803A (en) Programmable digital hearing aid
ES450719A1 (en) Arrangement for recognizing sounds
GB1591996A (en) Apparatus for recognising words from among continuous speech
US4817155A (en) Method and apparatus for speech analysis
US3821472A (en) Cough monitoring apparatus
CA1268822A (en) Systems and methods for monitoring the fetal heart
US3883850A (en) Programmable word recognition apparatus
US3855417A (en) Method and apparatus for phonation analysis lending to valid truth/lie decisions by spectral energy region comparison
US6285979B1 (en) Phoneme analyzer
US4157456A (en) Audiometer
US4219695A (en) Noise estimation system for use in speech analysis
KR102154776B1 (en) CCTV surveillance system that analyzes sound characteristics and outputs them with video
US3245403A (en) System for acoustic detection of pathologic larynges
US4401850A (en) Speech analysis apparatus
US3846586A (en) Single oral input real time analyzer with written print-out