DESC:TECHNICAL FIELD
[0001] The present invention relates generally to a system and method for impulse sound detection and classification. The present invention, more particularly, relates to a hybrid system and method for desired impulse sound detection and classification.

BACKGROUND
[0002] Impulse sound is a high energy level sound which is generally caused by an impact. Many conventional systems have been developed for detecting the impulse sound. For example, US10089845B2 describes a system and method of detecting and analyzing a threat in a confined environment. An audio board detects and analyzes audio signals. The audio board includes a microcontroller with at least one band-pass filter for distinguishing between a threat and a non-threat event and for measuring or counting pulses if the event is a threat.

[0003] US5455868A describes an amplitude detection system which analyzes the amplitude characteristic of a received noise and determines whether that characteristic conforms to the predictable audio signature of a gunshot. If a received noise reaches a predetermined amplitude level within a rise time that may be indicative of a gunshot, subsequent amplitude criteria are established representing the decay of the amplitude profile that is expected if the noise is a gunshot.

[0004] However, the conventional systems are limited to detect impulse sound of specific elements, for example gunshot, as well as other areas. Further, these conventional systems are not able to classify a type of the impulse sound, for example whether the generated impulse sound in an area is desired or non-desired impulse sound.

[0005] Therefore, there is a need of a system and method which solves the above defined problems and can detect the desired impulse sound in an open space as well as confined environment to reject multipath desired impulse sounds in classification. There is also a need of a system and method that can do better detection by considering frequency domain and time domain features compared to amplitude criteria.
SUMMARY
[0006] This summary is provided to introduce concepts related to a hybrid system and method for desired impulse sound detection and classification. This summary is neither intended to identify essential features of the present invention nor is it intended for use in determining or limiting the scope of the present invention.

[0007] For example, various embodiments herein may include one or more systems and methods for desired impulse sound detection and classification. In one of the embodiments, a method for detecting and classifying impulse sound includes a step of receiving, by a microphone, sound signals from one or more sources. The method includes a step of acquiring, by an acquisition unit, the signals and generating a plurality of samples. The method includes a step of storing, in a buffer module, the generated samples. The method includes a step of detecting, by a detection unit, impulse sound from the stored samples. The method includes a step of classifying, by a classification unit, the detected impulse sound based on one or more features.

[0008] In another embodiment, a system for detecting and classifying impulse sound includes a microphone, an acquisition unit, a buffer module, a detection unit, and a classification unit. The microphone is configured to receive sound signals from one or more sources. The acquisition unit is configured to acquire the signals and generate a plurality of samples. The buffer module is configured to store the generated samples. The detection unit is configured to detect impulse sound from the stored samples. The classification unit is configured to classify the detected impulse sound based on one or more features.

BRIEF DESCRIPTION OF ACCOMPANYING DRAWINGS
[0009] The detailed description is described with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The same numbers are used throughout the drawings to reference like features and modules.

[0010] Figure 1 illustrates a block diagram depicting a system for detecting and classifying impulse sound, according to an implementation of the present invention.

[0011] Figure 2 illustrates a flowchart depicting a method for detecting and classifying impulse sound, according to an exemplary implementation of the present invention.

[0012] Figure 3 illustrates a flowchart depicting a structural flow of impulse detection, according to an exemplary implementation of the present invention.

[0013] Figure 4 illustrates a flowchart depicting a structural flow of a temporal feature based classification unit, according to an exemplary implementation of the present invention.

[0014] Figure 5 illustrates a flowchart depicting a structural flow of a spectral feature based classification unit, according to an exemplary implementation of the present invention.

[0015] Figure 6A illustrates a graphical representation depicting a waveform of desired impulse, according to an exemplary implementation of the present invention.
[0016] Figure 6B illustrates a graphical representation depicting a waveform of clap, according to an exemplary implementation of the present invention.

[0017] Figure 6C illustrates a graphical representation depicting a waveform of tap on metal, according to an exemplary implementation of the present invention.

[0018] Figure 6D illustrates a graphical representation depicting a waveform of finger snapping, according to an exemplary implementation of the present invention.

[0019] Figure 6E illustrates a graphical representation depicting a waveform of a cracker, according to an exemplary implementation of the present invention.

[0020] Figure 6F illustrates a graphical representation depicting a waveform of a balloon, according to an exemplary implementation of the present invention.

[0021] It should be appreciated by those skilled in the art that any block diagrams herein represent conceptual views of illustrative systems embodying the principles of the present invention. Similarly, it will be appreciated that any flowcharts, flow diagrams, and the like represent various processes which may be substantially represented in computer readable medium and so executed by a computer or processor, whether or not such computer or processor is explicitly shown.
DETAILED DESCRIPTION
[0022] In the following description, for the purpose of explanation, specific details are set forth in order to provide an understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without these details. One skilled in the art will recognize that embodiments of the present invention, some of which are described below, may be incorporated into a number of systems.
[0023] The various embodiments of the present invention provide a hybrid system and method for desired impulse sound detection and classification. Furthermore, connections between components and/or modules within the figures are not intended to be limited to direct connections. Rather, these components and modules may be modified, re-formatted or otherwise changed by intermediary components and modules.

[0024] References in the present invention to “one embodiment” or “an embodiment” mean that a particular feature, structure, characteristic, or function described in connection with the embodiment is included in at least one embodiment of the invention. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment.

[0025] In one of the embodiments, a method for detecting and classifying impulse sound includes a step of receiving, by a microphone, sound signals from one or more sources. The method includes a step of acquiring, by an acquisition unit, the signals and generating a plurality of samples. The method includes a step of storing, in a buffer module, the generated samples. The method includes a step of detecting, by a detection unit, impulse sound from the stored samples. The method includes a step of classifying, by a classification unit, the detected impulse sound based on one or more features.

[0026] In another implementation, the method includes a step of storing the generated samples in a circular buffer.

[0027] In another implementation, the detected impulse sound is implemented on a Field Programmable Gate Array (FPGA).

[0028] In another implementation, the detected impulse sound is implemented on said FPGA by: capturing a plurality of block samples of present data; creating a first buffer corresponding to the circular buffer, wherein the first buffer is having first samples of present data; creating a second buffer corresponding to the circular buffer, wherein the second buffer is having second samples of pre-acquired data; calculating an absolute sum of the first and second samples within each first buffer and second buffer; checking whether absolute sum of the first buffer is greater than at least five times of an absolute sum of the second buffer; detecting the impulse sound based on the checked absolute sum; and capturing the plurality of samples from the detected impulse sound and storing in the buffer module for classification.

[0029] In another implementation, the method includes a step of classifying the detected impulse sound in time and frequency domains.

[0030] In another implementation, the method includes a step of classifying the detected impulse sound as desired impulse sound or non-desired impulse sound.

[0031] In another implementation, the one or more features include a spectral feature, a temporal feature, and a combination thereof.

[0032] In another implementation, the method includes a step of detecting impulse sound and classifying the impulse sound is performed in temporal and spectral domains.

[0033] In another implementation, the method includes a step of extracting, by an extractor, spectral and temporal features from the detected impulse sound.

[0034] In another implementation, the method includes a step of extracting the temporal features further includes a step of removing, by the extractor, an offset and normalizing the acquired signal. The method includes a step of smoothing of the normalized signal, by the extractor. The method includes a step of determining, by the extractor, peaks from the smoothing signals. The method includes a step of computing, by the extractor, the energy of the smoothing signals by thresholding with standard deviation of the smoothing signals. The method includes a step of identifying, by the extractor, one or more edges of an energy pulse. The method includes a step of estimating, by the extractor, features from the energy pulse determining zero crossings in a pre-determined time interval, a number of peaks between the zero crossings, and periodicity of the smoothing signals.

[0035] In another implementation, the method includes a step of extracting the spectral features includes a step of removing, by the extractor, an offset and normalizing the acquired signal. The method includes a step of computing, by the extractor, the Fast Fourier Transform of the normalized signal for estimating spectral features. The method includes a step of passing through a pre-estimated triangular filter bank with uniformly spaced filters. The method includes a step of applying a cosine transform on a signal corresponding to individual bands for identifying Cepstral coefficients as spectral features.

[0036] In another implementation, the method includes a step of classifying the detected impulse sound based on spectral features and includes a step of recording desired impulse sounds by using a high sound pressure level (SPL) microphone. The method includes a step of extracting the spectral features for all the recorded desired impulses. The method includes a step of generating a Gaussian mixture model using the spectral features and estimating the mean and covariance. The method includes a step of estimating an error using Mahalanobis distance between the spectral features and the Gaussian mixture model.

[0037] In another embodiment, a system for detecting and classifying impulse sound includes a microphone, an acquisition unit, a buffer module, a detection unit, and a classification unit. The microphone is configured to receive sound signals from one or more sources. The acquisition unit is configured to acquire the signals and generate a plurality of samples. The buffer module is configured to store the generated samples. The detection unit is configured to detect impulse sound from the stored samples. The classification unit is configured to classify the detected impulse sound based on one or more features.

[0038] In an exemplary embodiment, the hardware is changed to the system on a chip configuration by replacing the FPGA and the classification unit.

[0039] In an exemplary embodiment, an analog microphone and an analog-to -digital (A/D) converter is replaced with a high sound pressure level (SPL) Micro Electrical-Mechanical (MEMs) microphone.

[0040] In an exemplary embodiment, the system and method are used in localization of desired impulse with multiple arrays of microphones.

[0041] In an exemplary embodiment, after detecting threat, the information is communicated through wired or wireless communication.

[0042] In an exemplary embodiment, the spectral feature based classification unit is trained with impulse sounds for surveillance applications.

[0043] Figure 1 illustrates a block diagram depicting a system(100) for detecting and classifying impulse sound, according to an implementation of the present invention.

[0044] A system for detecting and classifying impulse sound (hereinafter referred to as “system”) (100) includes a microphone (102), an acquisition unit (104), a buffer module (106), a detection unit (108), and a classification unit (110).

[0045] The microphone (102) is deployed over an area to monitor ambient acoustic signals. In an embodiment, the system (100) includes a plurality of microphones which are arranged in an array and deployed to cover a large area. The microphone (102) is configured to receive the signals from one or more sources. In an embodiment, the sources can include any area where acoustic sound is generated. The area can include a confined space or an open space. The microphone (102) includes an Analog to Digital (A/D) Converter (not shown in a figure) which is configured to convert the received signal into digital signals. In an embodiment, the microphone (102) can be a high sound pressure level (SPL) microphone.

[0046] The acquisition unit (104) is configured to cooperate with the microphone to receive the digital signals. The acquisition unit (104) is further configured to acquire the digital signals and generate a plurality of samples.

[0047] The buffer module (106) is configured to cooperate with the acquisition unit (104) to receive and store the generated samples. In an embodiment, the buffer module (106) is configured to store the generated samples in a circular buffer.

[0048] The detection unit (108) is configured to cooperate with the buffer module (106) to receive the stored samples. The detection unit (108) is further configured to detect impulse sound from the stored samples. In an embodiment, the detected impulse sound is implemented on a Field Programmable Gate Array (FPGA). In another embodiment, for implementing the detected impulse sound on the FPGA, the detection unit (108) is configured to capture a plurality of block samples of present data. The detection unit (108) is further configured to create a first buffer and a second buffer corresponding to the circular buffer. The first buffer is having first samples of present data and the second buffer is having second samples of pre-acquired data. The detection unit (108) is further configured to calculate an absolute sum of the first and second samples within each first buffer and second buffer and check whether an absolute sum of the first buffer is greater than at least five times of an absolute sum of the second buffer. Thereafter, the detection unit (108) is configured to detect the impulse sound based on the checked absolute sum and capture the plurality of samples from the detected impulse sound and store in the buffer module (106) for classification. In an embodiment, the detection unit (108) is configured to detect impulse sound in temporal and spectral domains.

[0049] In an exemplary embodiment, the detection unit (108) is further configured to create a first buffer and a second buffer. The first buffer is having 128 samples of present data and the second buffer is having 128 samples of previous data. All the samples in each buffer are added and a check is performed to observe if the present sample buffer sum is greater than at least five times of previous samples buffer sum for detection of impulse. After detection of impulse, total 10240 samples are captured and saved into the memory for classification.

[0050] The classification unit (110) is configured to cooperate with the detection unit (108) to receive the detected impulse sound. The classification unit (110) is further configured to classify the detected impulse sound based on one or more features. In an embodiment, the one or more features include, but are not limited to, a spectral feature, a temporal feature, or any combinations thereof. In an embodiment, the classification unit (110) is configured to classify the detected impulse sound in time and frequency domains. In another embodiment, the classification unit (110) is configured to classify the detected impulse sound as a desired impulse sound or a non-desired impulse sound. In an embodiment, the classification unit (110) is configured to classify the detected impulse sound in temporal and spectral domains.

[0051] In an embodiment, the classification unit (110) includes a time domain classifier (112) and a frequency domain classifier (114). The time domain classifier (112) is configured to classify the detected impulse sound based on the time domain features. The frequency domain classifier (114) is configured to classify the detected impulse sound based on the frequency domain features. In one embodiment, the time domain classifier (112) and the frequency domain classifier (114) are configured to distinguish whether the captured impulse sound is desired impulse or non-desired impulse.

[0052] In an embodiment, the system (100) includes an extractor (116). The extractor (116) is configured to extract spectral and temporal features from the detected impulse sound for further classification.

[0053] In an embodiment, for extracting the temporal features, the extractor (116) is configured to remove an offset and normalize the acquired signal received from the acquisition unit (104). The extractor (116) is further configured to smooth the normalized signal and determine peaks from the smoothing signals. Further, the extractor (116) is configured to compute the energy of the smoothing signals by thresholding with standard deviation of the smoothing signals, identify one or more edges of an energy pulse, and estimate features from the energy pulse determining zero crossings in a pre-determined time interval, a number of peaks between the zero crossings, and periodicity of the smoothing signals.

[0054] In an exemplary embodiment, the temporal features are extracted by removing a DC offset and normalizing the signal captured in the memory, smoothing the signal for getting proper peaks, computing the energy of the signal and thresholding with a standard deviation of the signal, resulting in the estimation of the leading edge (t1) and trailing edge (t2) of the energy pulse. In another exemplary embodiment, temporal features are estimated by finding zero crossings in the time duration [t1, t2], a number of peaks between the zero crossings in the time duration [t1, t2], and the periodicity of the signal.

[0055] In an embodiment, for extracting the spectral features, the extractor (116) is configured to remove an offset and normalize the acquired signal. The extractor (116) is further configured to compute the Fast Fourier Transform of the normalized signal for estimating spectral features and pass through a pre-estimated triangular filter bank with uniformly spaced filters. The extractor (116) then applies a cosine transform on a signal corresponding to individual bands for identifying Cepstral coefficients as spectral features.

[0056] In an exemplary embodiment, the spectral features are extracted by removing the DC offset and normalizing the signal captured in the memory, computing the Fast Fourier Transform of the signal to estimate spectral features, passing through a pre-estimated triangular filter bank with uniformly spaced filters on equal scale from zero to 10,000 Hz frequency, and applying the cosine transform on the signal corresponding to the individual bands to get the Cepstral coefficients as spectral features.

[0057] In an embodiment, the classification unit (110) is configured to classify the detected impulse sound based on spectral features by recording desired impulse sounds by using a high sound pressure level (SPL) microphone, extracting the spectral features for all the recorded desired impulses, generating a Gaussian mixture model using the spectral features and estimating the mean and covariance, and estimating an error using Mahalanobis distance between the spectral features and the Gaussian mixture model.

[0058] In an exemplary embodiment, the acquisition unit (104) acquires desired impulse samples. These samples are then stored in a buffer module (106). Thereafter, the system (100) raises a memory flag in the buffer module (106) and starts processing. The stored samples are used for detecting and classifying the impulse sound. The detection module (108) detects the energy of the signal based on dynamic adaptive threshold. The classification unit (110) classifies the detected impulse sounds as desired impulse or non-desired impulse based on multiple spectral and temporal features extracted from the detected impulse sounds and by multiple ways.

[0059] In an exemplary embodiment, the system (100) includes a controller (not shown in a figure) which is configured to control various functionalities of the classification unit (110). In an embodiment, the classification unit (110) runs on the controller.

[0060] Figure 2 illustrates a flowchart (200) depicting a method for detecting and classifying impulse sound, according to an exemplary implementation of the present invention.

[0061] In Figure 2, the flowchart (200) starts at a step (202), receiving, by a microphone, sound signals from one or more sources. In an embodiment, a microphone (102) is configured to receive sound signals from one or more sources. At a step (204), acquiring, by an acquisition unit, the signals and generating a plurality of samples. In an embodiment, an acquisition unit (104) is configured to acquire the signals and generating a plurality of samples. At a step (206), storing, in a buffer module, the generated samples. In an embodiment, a buffer module (106) is configured to store the generated samples. At a step (208), detecting, by a detection unit, impulse sound from the stored samples. In an embodiment, a detection unit (108) is configured to detect impulse sound from the stored samples. At a step (210), classifying, by a classification unit, the detected impulse sound based on one or more features. In an embodiment, the classification unit (110) is configured to classify the detected impulse sound based on one or more features.

[0062] Figure 3 illustrates a flowchart (300) depicting a structural flow of impulse detection, according to an exemplary implementation of the present invention.

[0063] In Figure 3, the flowchart (300) starts at a step (302), acquiring microphone data from an analog to digital converter. In an embodiment, an acquiring unit (104) is configured to acquire the microphone data. In an embodiment, the microphone data is in digitized form. In an exemplary embodiment, the microphone data is digitized with 62.5 KHz sampling rate. At a step (304), storing the data as a block in a circular buffer. In an embodiment, the buffering module (106) is configured to store the data as a block in a circular buffer, for further processing. In an exemplary embodiment, the block size is 128. At a step (306), calculating an absolute sum of the samples in each block. In an embodiment, to estimate energy in the rise time of impulse, an absolute sum of each block has been taken and compared present block with the previous block. If the absolute sum in the present block is greater than five times of a previous block, then the signal is impulse in nature because of the high rise time of the signal (as shown at a block (308)). Once the impulse is detected, 10240 samples from the analog to digital converter are captured for further analysis, as shown at a step (310). Otherwise the steps (302), (304), (306), and (308) are repeated for impulse detection.

[0064] Figure 4 illustrates a flowchart (400) depicting a structural flow of a temporal feature based classification unit (110), according to an exemplary implementation of the present invention.

[0065] In Figure 4, the flowchart starts at a step (402), removing the DC offset and normalizing the captured signal. In an embodiment, the extractor (116) is configured to remove the DC offset and normalize the captured signal. In an embodiment, the normalization of the signal is done by dividing the signal captured in the step (310) of the Figure 3 with an absolute maximum sample value. At a step (404), filtering the signal for smoothing. In an embodiment, the extractor (116) is configured to filter the signal for smoothing. In one embodiment, a digital low pass Butterworth filter or order 3 with cutoff frequency 2 kHz is implemented to smooth the signal. At a step (406), estimating the energy of the signal with a block size of 100. In an embodiment, the extractor (116) is configured to estimate the energy of the signal with a block size of 100. The energy signal is estimated by performing block processing of the signal of block size 100. At a step (408), determining the start and end of the event by using energy of the signal which is greater than standard deviation of the signal. In an embodiment, the extractor (116) is configured to determine the start and end of the event by using energy of the signal which is greater than standard deviation of the signal. In one embodiment, start of the event is estimated where the energy signal is crossing threshold and end of the event is estimated by circularly shifting the energy signal and applying threshold. The estimated event is taken for analyzing time domain features of the sound signal. At a step (410), generating a signal with peaks and valleys within an event. In an embodiment, the extractor (116) is configured to generate a signal with peaks and valleys within an event. In an embodiment, the extractor (116) generates a signal having only peaks and valleys within the start and end of the event. At a step (412), estimating the temporal features zero crossing, periodicity, rise time of initial peak, and fall time of a first valley. In an embodiment, by keeping thresholds in temporal features on the signal generated in the step (410), the classification unit (110) can classify the impulse signal as desired impulse or not. Once it passes the step (412), the signal in the step (310) is fed to the Frequency domain classifier (114).

[0066] Figure 5 illustrates a flowchart (500) depicting a structural flow of a spectral feature based classification unit, according to an exemplary implementation of the present invention.

[0067] In Figure 5, the flowchart (500) starts at a step (502), removing the DC off set and normalizing the captured signal. In an embodiment, the extractor (116) is configured to remove the DC offset and normalize the captured signal. In an embodiment, the DC offset correction is done by removing mean of the captured signal and the normalization of the signal is done by dividing the signal captured in the step (310) of the Figure 3 with an absolute maximum sample value. At a step (504), grouping of 0 to 10 KHz frequency into triangular bands of equal bandwidth. In an embodiment, the extractor (116) is configured to group 0 to 10 KHz frequency into triangular bands of equal bandwidth. In one embodiment, generation of a triangular filter bank matrix of each row is a triangular filter. In the step (504), it groups 0 to 10 KHz frequency range into 20 uniformly spaced triangular bands of equal bandwidth. At a step (506), applying Fast Fourier Transform on uniformly spaced Triangular filter bands. In an embodiment, the extractor (116) is configured to apply Fast Fourier Transform on uniformly spaced Triangular filter bands. In an embodiment, the step (506) illustrates a filter bank application as a unique part of the magnitude spectrum. The magnitude spectrum is calculated by applying Fast Fourier Transform on the signal generated at the step (502). At a step (508), finding the discrete cosine transforms computation of bands. In an embodiment, the extractor (116) is configured to find the discrete cosine transforms computation of bands. In one embodiment, the extractor (116) is configured to estimate the cepstral coefficients through the discrete cosine transforms. These cepstral coefficients generated at the step (508) are used as spectral features for the classification unit (110). A Gaussian Mixture Model classifier is used for classification with spectral features estimated at the step (508). At a step (510), storing the DCT coefficients as spectral features for training and classifying desired impulse sound. In an embodiment, the buffer module (106) is configured to store the DCT coefficients as spectral features for training and classifying desired impulse sound. At a step (512), estimating the spectral features for all recorded desired impulse sounds. In an embodiment, the classification unit (110) is configured to estimate the spectral features for all recorded desired impulse sounds. In one embodiment, real desired impulse sounds are recorded at all trial sites and are stored for training and all the recorded desired impulses spectral features are estimated. At a step (514), fitting a Gaussian mixture distribution model to spectral features of all recorded desired impulse sounds. In an embodiment, the classification unit (110) is configured to fit a Gaussian mixture distribution model to spectral features of all recorded desired impulse sounds. In one embodiment, a Gaussian mixture distribution model is fitted to spectral features of all recorded desired impulse sounds. At a step (516), storing the covariance matrix and a mean vector of a Gaussian mixture distribution model into a memory. In an embodiment, the buffer module (106) is configured to store the covariance matrix and a mean vector of a Gaussian mixture distribution model into a memory. In one embodiment, the buffer module (106) is configured to save the covariance and the mean vector of the generated GMM model into the memory which are used for classifying live events captured from the microphone (102). At a step (518), finding Mahalanobis distance between spectral features of the event and the Gaussian mixture distribution model. In an embodiment, the classification unit (110) is configured to find Mahalanobis distance between spectral features of the event and the Gaussian mixture distribution model. Once the signal passes to the time domain classifier (112) of the classification unit (110) (as shown in Figure 1), the captured signal of the step (310) as shown at the Figure 3 is passed through the steps (502), (504), (506), (508), and (518). If the Mahalanobis distance is less than the error fixed in the classification unit (110), (as shown at a step (520), for the desired impulse, then the sound is classified as a desired impulse sound, as shown at a step (522). If the distance is more than the error, then the sound is non-desired impulse sound, as shown at a step (524).

[0068] Figure 6A illustrates a graphical representation depicting a waveform of desired impulse, according to an exemplary implementation of the present invention. More specifically, Figure 6A illustrates a desired impulse waveform recorded from the microphone (102).

[0069] Figure 6B illustrates a graphical representation depicting a waveform of clap, according to an exemplary implementation of the present invention. More specifically, Figure 6B illustrates a clap waveform recorded from the microphone (102) by clapping hands that produces impulses.

[0070] Figure 6C illustrates a graphical representation depicting a waveform of tap on metal, according to an exemplary implementation of the present invention. More specifically, Figure 6C illustrates a tap waveform recorded from the microphone (102) by tapping on a metal body that produces impulses.
[0071] Figure 6D illustrates a graphical representation depicting a waveform of finger snapping, according to an exemplary implementation of the present invention. More specifically, Figure 6D illustrates a waveform recorded from the microphone (102) by snapping fingers that produces impulses.

[0072] Figure 6E illustrates a graphical representation depicting a waveform of a cracker, according to an exemplary implementation of the present invention. More specifically, Figure 6E illustrates a waveform recorded from the microphone (102) by a blasting cracker that produces impulses.

[0073] Figure 6F illustrates a graphical representation depicting a waveform of a balloon, according to an exemplary implementation of the present invention. More specifically, Figure 6F illustrates a waveform recorded from the microphone (102) by a blasting balloon that produces impulses.

[0074] It should be noted that the description merely illustrates the principles of the present invention. It will thus be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described herein, embody the principles of the present invention. Furthermore, all examples recited herein are principally intended expressly to be only for explanatory purposes to help the reader in understanding the principles of the invention and the concepts contributed by the inventor(s) to furthering the art and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the invention, as well as specific examples thereof, are intended to encompass equivalents thereof.

,CLAIMS:
1. A method for detecting and classifying impulse sound, comprising:
receiving, by a microphone (102), sound signals from one or more sources;
acquiring, by an acquisition unit (104), said signals and generating a plurality of samples;
storing, in a buffer module (106), said generated samples;
detecting, by a detection unit (108), impulse sound from said stored samples; and
classifying, by a classification unit (110), said detected impulse sound based on one or more features.

2. The method as claimed in claim 1, wherein storing said generated samples in a circular buffer.

3. The method as claimed in claim 1, wherein said detected impulse sound is implemented on a Field Programmable Gate Array (FPGA).

4. The method as claimed in claim 3, wherein said detected impulse sound is implemented on said FPGA by:
capturing a plurality of block samples of present data;
creating a first buffer corresponding to said circular buffer, wherein said first buffer is having first samples of present data;
creating a second buffer corresponding to said circular buffer, wherein said second buffer is having second samples of pre-acquired data;
calculating an absolute sum of said first and second samples within each first buffer and second buffer;
checking whether an absolute sum of said first buffer is greater than at least five times of an absolute sum of said second buffer;
detecting said impulse sound based on said checked absolute sum; and
capturing said plurality of samples from said detected impulse sound and storing in said buffer module for classification.

5. The method as claimed in claim 1, wherein classifying said detected impulse sound in time and frequency domains.

6. The method as claimed in claim 1, wherein classifying said detected impulse sound as desired impulse sound or non-desired impulse sound.

7. The method as claimed in claim 1, wherein said one or more features include a spectral feature, a temporal feature, and combinations thereof.

8. The method as claimed in claim 1, wherein said steps of detecting impulse sound and classifying said impulse sound are performed in temporal and spectral domains.

9. The method as claimed in claim 8, comprising extracting, by an extractor (116), spectral and temporal features from said detected impulse sound.

10. The system as claimed in claim 9, wherein extracting said temporal features comprising:
removing, by said extractor (116), an offset and normalizing said acquired signal;
smoothing, by said extractor (116), said normalized signal;
determining, by said extractor (116), peaks from said smoothing signals;
computing, by said extractor (116), the energy of said smoothing signals by thresholding with standard deviation of said smoothing signals;
identifying, by said extractor (116), one or more edges of an energy pulse; and
estimating, by said extractor (116), features from said energy pulse determining zero crossings in a pre-defined time interval, a number of peaks between the zero crossings, and periodicity of said smoothing signals.

11. The method as claimed in claim 9, wherein extracting said spectral features comprising:
removing, by said extractor (116), an offset and normalizing said acquired signal;
computing, by said extractor (116), the Fast Fourier Transform of said normalized signal for estimating spectral features;
passing through a pre-estimated triangular filter bank with uniformly spaced filters; and
applying a cosine transform on a signal corresponding to individual bands for identifying Cepstral coefficients as spectral features.

12. The method as claimed in claim 9, wherein classifying said detected impulse sound based on spectral features comprising:
recording desired impulse sounds by using high sound pressure level (SPL) microphone (102);
extracting the spectral features for all the recorded desired impulses;
generating a Gaussian mixture model using the spectral features and estimating the mean and covariance; and
estimating an error using Mahalanobis distance between said spectral features and said Gaussian mixture model.

13. A system (100) for detecting and classifying impulse sound, said system comprising:
a microphone (102) configured to receive sound signals from one or more sources;
an acquisition unit (104) configured to cooperate with said microphone (102), said acquisition unit (104) configured to acquire said signals and generate a plurality of samples;
a buffer module (106) configured to cooperate with said acquisition unit (104), said buffer module (106) configured to store said generated samples;
a detection unit (108) configured to cooperate with said buffering module (106), said detection unit (108) configured to detect impulse sound from said stored samples; and
a classification unit (110) configured to cooperate with said detection unit (108), said classification unit (110) configured to classify said detected impulse sound based on one or more features.