Claims:WE CLAIM:

1. A percussion sound analysis system for detection of various abdominal conditions in non-invasive manner comprising

a sound acquisition system to capture the percussion sound signals from abdominal area of a subject;

a computing unit to process and analyze the percussion sound signals including pre-processor of signals, feature extractor and classifier enabling detection of abdominal conditions in the subject;

a cooperative user interface.

2. A percussion sound analysis system as claimed in claim 1 for detection of various abdominal conditions comprising

said sound acquisition system to capture the percussion sound signals from abdominal area of a subject;

said computing unit to process and analyze the percussion sound signals by executing preprocessing the signals, extracting features of the signals and classifying the features based on trained classifier or model in order detect abdominal conditions including related ailments in the subject;

a user interface to show the analyzed percussion sound signals and/or detected abdominal ailments in the subject.

3. A percussion sound analysis system as claimed in anyone of claim 1 or 2, wherein the sound acquisition system includes

acoustics medical device to pickup the percussion sound signals from the subject generated upon manual tapping over the abdominal area of the subject;

signal conditioner unit to receive the captured percussion sound signals and pre-emphasizing the signals comprising
pre-amplifier to amplify the captured Percussion sound signals at 0-2 kHz frequencies
filter to enhance the Percussion soundsignals with frequencies in the range of 0-2 kHz simultaneously suppress high frequency (above 2 kHz) noise components;
instrumentation amplifier to facilitate impedance matching with subsequent measurement and test equipment.

4. A percussion sound analysis system as claimed in claim 3, wherein the signal conditioner unit feeds the emphasized signals to a subsequent analog to digital converter and the computing unit having a Digital computing device and a Digital signal processing unit for analyzing the signals.

5. A percussion sound analysis system as claimed in anyone claims 1 to 4, wherein the computing unit comprises

Preprocessing unit to generate filtered normalized and decimated segmentation of the Percussion sound signals;

Feature extraction unit to extract general features of the segmented Percussion sound signal including maximum peak to peak amplitude, peak frequency in the segment, total power in each segment, mid-power frequency (frequency below which half of the total power is contained) and BMI of the subject to account for the varying thickness of skin tissues across the subject;

Random Forest Classifier to detect abdominal ailments in the subject by analyzing extracted the feature variables with respect to sampled training models according to objective determination of the abdominal ailments and involving a standard decision tree.

6. A percussion sound analysis system as claimed in anyone claims 1 to 5, wherein the Preprocessing unit segment the percussion sound signal ensuring that each segment of the signal contain exactly one percussion by first finding peaks of the signal to identify regions of percussion blows and then from the position of each peak, a marker is moved sufficiently to left side to ensure start of the signal and then moving the marker to the right and simultaneously measuring slope of the signal to identify onset of the percussion signal by using a threshold slope value and upon identification of the onset, a fixed length of the signal ( 10,000 samples ~ 225ms ) starting from the onset point is segmented.

7. A percussion sound analysis system as claimed in claim 6, wherein the Preprocessing unit further includes

sampler to decimate the segmented signal by a factor of 20 to reduce the signal sampling frequency to 2205 Hz and length to 500 ms.

normalization unit to scale each segment to have maximum amplitude of ±1 in order to nullify intensity variation in the percussion blows occurring due to manual tapping of the abdomen;

second order Butterworth band pass filter with cutoff frequencies 20 Hz and 1 kHz to filter the segmented percussion signal in order to eliminate the noise.

8. A percussion sound analysis system as claimed in anyone of claims 1 to 7, wherein the Feature extraction unit includes an assembly of second order band pass filters to obtain peak frequencies corresponding to frequency spectrum of the segmented signal by approximating the segmented signal as an exponentially decaying sinusoid after dividing it into various frequency sub-bands by filtering it through the filters of the assembly;
said assembly includes a second order band pass filter with cutoff frequencies 10 Hz and 230 Hz, a second order band pass filter with cutoff frequencies 220 Hz & 420 Hz, a second order band pass filter with cutoff frequencies 410 Hz & 560 Hz and a second order band pass filter with cutoff frequencies 550 Hz & 1000 Hz and major frequency components in each of the filtered signals provides features of the segmented signal.

9. A percussion sound analysis system as claimed in anyone of claims 5 to 8, wherein the Random Forest Classifier involves the standard decision trees in combination with notion of an ensemble;
said Random Forest Classifier analyze the Percussion sound signal features by
sampling 2/3rd of the training cases at random with replacement from the training model to create a subset of the data;
selecting m feature variables at random from all the feature variables at each node and involving the feature variable that provides best split according to objective determination of the abdominal ailments to do a binary split on that node;
choosing another m variable from all predictor variables performing same operation at next node;
outputting analysis result based on voting majority (when decision threshold is set as 0.5) of all of terminal nodes in the decision trees that are reached.

10. A percussion sound analysis system as claimed in anyone of claims 1 to 9, wherein the user interface includes

an auditory unit to play the captured percussion sound; and

a visual display unit to show the result of analysis of the trained classifier or the extracted feature metrics unit with those observed in past for different states of abdomen using a parallel co-ordinates graph.

11. A percussion sound analysis system as claimed in anyone of claims 1 to 9, wherein the training model to detect abdominal conditions including ailments in the subject comprise recorded sound of tapping in the abdominal area of volunteers with a neurosurgical hammer and processed each said recorded signal through the Preprocessing unit and Feature extraction unit;

said random forest classifier based on extracted features of the percussion sound signals of the volunteers enabling identifying the state of the abdomen for the subject.

12. A kit for percussion sound analysis comprising:

A) a percussion sound analysis system for detection of various abdominal conditions comprising (i) a sound acquisition system to capture the percussion sound signals from abdominal area of a subject; (ii) a computing unit to process and analyze the percussion sound signals including preprocessor of signals, feature extractor and classifier enabling detection of abdominal conditions in the subject; and (iii) a cooperative user interface; and
B) a neurosurgical hammer.
, Description:
FIELD OF THE INVENTION:
The present invention relates Percussion Sound Analysis technique in order to discriminate between different states of the underlying percussed organ. In particular, the present invention is directed to develop a non-invasive Percussion Sound Analysis system for identifying various abdominal ailments that a subject may suffer from. The present Percussion Sound Analysis system advantageously incorporates data acquisition, data display, data auditory, subject database and signal analysis to classify visually with images of features of the percussion sound acquired from abdominal region of the subject.

BACKGROUND ART.:
Percussion Sound Analysis is a technique where percussion sounds are used in order to discriminate between different states of the underlying percussed organ. Information indicating different states of the underlying percussed organ is extracted from the percussion sound signals by extracting various featuresof the percussion sound signals. Prevailing works in the Percussion Sound Analysis technique are mostly involved in generation and acquisition of the percussion signals.
A.B. Bohadana et al. [Ref.: A.B. Bohadana, S.S. Kraman, "Consistency of sternal percussion performed manually and with mechanical thumper”, European Respiratory Journal, Vol. 5, No. 8, 1004–1008, September 1992.] teaches about finger percussion which is used in study of acoustic behaviour of chest. The authors calculated the intra-subject variability and short-term reproducibility of finger percussion in 10 healthy subjects. They examined several indices of the output sound of two series of sternal percussion manoeuvres performed one hour apart by the same examiner. The results were compared to those obtained during sternal percussion performed by a mechanical thumper. They concluded that finger percussion of the sternum is sufficiently consistent to be used as a tool to investigate the acoustic behaviour of the chest.
H.A. Mansy et al. [Ref.:H.A. Mansy, T. J. Royston and R. H. Sandler, "Use of abdominal percussion for pneumoperitoneum detection”, Medical & Biological Engineering & Computing, Vol. 40, No. 4, 439–446, July 2002.] discloses about diagnostic changes in the sounds induced by abdominal percussion resulting from abdominal structure alterations due to pneumoperitoneum. The study investigated these changes in a mongrel dog model. Abdominal percussion was performed at baseline and after creation of pneumoperitoneum states. The results suggested that the normal and the 1,000 ml pneumoperitoneum states can be separated, but separating the two levels of pneumoperitoneum was not feasible using the proposed approach. The authors concluded that analysis of abdominal percussion sounds may prove useful for pneumoperitoneum detection, but not for distinguishing different levels of that condition.
Bhuiyan Md. [Ref.: Bhuiyan Md. Moinuddin, "Automated Classification of Medical Percussion Signals for the Diagnosis of Pulmonary Injuries”, Electronic Theses And Dissertations, Paper 4941, 2013] discusses about various techniques for percussion sound analysis.
Moinuddin Bhuiyan, et al. [Ref.: Moinuddin Bhuiyan, Eugene V. Malyarenko, Mircea A. Pantea, Dante Capaldi, Alfred E. Baylor, and Roman Gr. Maev. “Time-Frequency Analysis of Clinical Percussion Signals Using Matrix Pencil Method”, Journal Of Electrical And Computer Engineering, Vol. 2015, Article Id 274541, 10 Pages, 2015.] discusses about time-frequency analysis of clinical percussion signals produced by tapping over human chest or abdomen with a neurological hammer and recorded with an air microphone. Matrix Pencil Method (MPM) is used to decompose the signal into a set of exponentially damped sinusoids, which are then plotted in the time-frequency plane. Such representation provides better visualization of the signal structure than the commonly used frequency-amplitude plots and facilitates tracking subtle changes in the signal for diagnostic purposes.
US Patent 4196722 teaches about a hand-held, manually operated percussion instrument for using in respiratory therapy. The instrument can only collect percussion sounds it cannot analyses the percussion sound.
US Patent 8160877 discloses an apparatus for acoustic percussion of a body. This apparatus also adapted to collect percussion sounds not analyze the sound.
Thus, the existing percussion sound based body ailments detection technique do not teaches about implementing the body percussion sound analysis technique in order to identify various abdominal ailments that a subject may suffer from. Further, there is no percussion sound analysis methodology reported in the art which can use data acquisition, data display, data auditory, subject database and signal analysis technique in combination to make the ailments detection technique much easier, so that the body ailments can be determined without the help of a trained medical professional. Hence, there has been a need for developing a percussion sound system which will accurately determine various abdominal ailments that a subject may suffer from based on the percussion sound acquired from the subject and should be easy to operate.

OBJECTIVE OF THE INVENTION:
It is thus the basic objective of the present invention is to develop a percussion sound analysis system, which would be adapted to analyze percussion sounds from the abdominal region of a subject in order to identify various abdominal ailments that the subject may suffer from.

Another objective of the present invention is to develop a percussion sound analysis system, which would be adapted to involve data acquisition and analysis technique to automatically determine the abdominal ailments based on a model of previous variation in the percussion sound in accordance with different abdominal status.

Another objective of the present invention is to develop a percussion sound analysis system to determine various abdominal ailments in a subject, which would be adapted to provide result of determining various abdominal ailments in the subject verified/determined based on machine learning based suggestion and/or visual analysis.

Another objective of the present invention is to develop a percussion sound analysis system for analyzing the percussion sounds and determining the abdominal ailments, which would be adapted to operate without relying on experience of a doctor and able to overcome the natural limitations of a human ear.

Yet another objective of the present invention is to develop a percussion sound analysis system for detection of various abdominal conditions in non-invasive manner, which would be cheap, small in size and easy to operate so that it can be operated by any person without having any medical training.

SUMMARY OF THE INVENTION:
Thus according to the basic aspect of the present invention, there is provided a percussion sound analysis system for detection of various abdominal conditions in non-invasive manner comprising

a sound acquisition system to capture the percussion sound signals from abdominal area of a subject;

a computing unit to process and analyze the percussion sound signals including pre-processor of signals, feature extractor and classifier enabling detection of abdominal conditions in the subject;

a cooperative user interface.

According to another aspect in the present invention there is provided a percussion sound analysis system for detection of various abdominal conditions comprising

said sound acquisition system to capture the percussion sound signals from abdominal area of a subject;

said computing unit to process and analyze the percussion sound signals by executing preprocessing the signals, extracting features of the signals and classifying the features based on trained classifier or model in order detect abdominal conditions including related ailments in the subject;

a user interface to show the analyzed percussion sound signals and/or detected abdominal ailments in the subject.

According to yet another aspect in the present percussion sound analysis system, the sound acquisition system includes

acoustics medical device to pickup the percussion sound signals from the subject generated upon manual tapping over the abdominal area of the subject;

signal conditioner unit to receive the captured percussion sound signals and pre-emphasizing the signals comprising
pre-amplifier to amplify the captured Percussion sound signals at 0-2 kHz frequencies
filter to enhance the Percussion soundsignals with frequencies in the range of 0-2 kHz simultaneously suppress high frequency (above 2 kHz) noise components;
instrumentation amplifier to facilitate impedance matching with subsequent measurement and test equipment.

According to another aspect in the present percussion sound analysis system, the signal conditioner unit feeds the emphasized signals to a subsequent analog to digital converter and the computing unit having a Digital computing device and a Digital signal processing unit for analyzing the signals.

According to another aspect in the present percussion sound analysis system, the computing unit comprises

Preprocessing unit to generate filtered normalized and decimated segmentation of the Percussion sound signals;

Feature extraction unit to extract general features of the segmented Percussion sound signal includingmaximum peak to peak amplitude, peak frequency in the segment, total power in each segment, mid-power frequency (frequency below which half of the total power is contained) and BMI of the subject to account for the varying thickness of skin tissues across the subject;

Random Forest Classifier to detect abdominal ailments in the subject by analyzing extracted the feature variables with respect to sampled training models according to objective determination of the abdominal ailments and involving a standard decision tree.

According to yet another aspect in the present percussion sound analysis system, the Preprocessing unit segment the percussion sound signal ensuring that each segment of the signal contain exactly one percussion by first finding peaks of the signal to identify regions of percussion blows and then from the position of each peak, a marker is moved sufficiently to left side to ensure start of the signal and then moving the marker to the right and simultaneously measuring slope of the signal to identify onset of the percussion signal by using a threshold slope value and upon identification of the onset, a fixed length of the signal ( 10,000 samples ~ 225ms ) starting from the onset point is segmented.

According to a further aspect in the present percussion sound analysis system, the Preprocessing unit further includes

sampler to decimate the segmented signal by a factor of 20 to reduce the signal sampling frequency to 2205 Hz and length to 500 ms.

normalization unit to scale each segment to have maximum amplitude of ±1 in order to nullify intensity variation in the percussion blows occurring due to manual tapping of the abdomen;

second order Butterworth band pass filter with cutoff frequencies 20 Hz and 1 kHz to filter the segmented percussion signal in order to eliminate the noise.

According to yet another aspect in the present percussion sound analysis system, the Feature extraction unit includes an assembly of second order band pass filters to obtain peak frequencies corresponding to frequency spectrum of the segmented signal by approximating the segmented signal as an exponentially decaying sinusoid after dividing it into various frequency sub-bands by filtering it through the filters of the assembly;
said assembly includes a second order band pass filter with cutoff frequencies 10 Hz and 230 Hz, a second order band pass filter with cutoff frequencies 220 Hz & 420 Hz, a second order band pass filter with cutoff frequencies 410 Hz & 560 Hz and a second order band pass filter with cutoff frequencies 550 Hz& 1000 Hz and major frequency components in each of the filtered signals provides features of the segmented signal.

According to yet another aspect in the present percussion sound analysis, the Random Forest Classifier involves the standard decision trees in combination with notion of an ensemble;
said Random Forest Classifier analyze the Percussion sound signal features by
sampling 2/3rd of the training cases at random with replacement from the training model to create a subset of the data;
selecting m feature variablesat random from all the feature variables at each node and involving the feature variable that provides best splitaccording to objective determination of the abdominal ailments to do a binary split on that node;
choosing another m variable from all predictor variables performing same operation at next node;
outputting analysis result based on voting majority (when decision threshold is set as 0.5) of all of terminal nodes in the decision trees that are reached.

According to yet another aspect in the present percussion sound analysis system, the user interface includes

an auditory unit to play the captured percussion sound; and

a visual display unit to show the result of analysis of the trained classifier or the extracted feature metric sunit with those observed in past for different states of abdomen using a parallel co-ordinates graph.

According to another aspect in the present percussion sound analysis system, the training model to detect abdominal conditions including ailments in the subject comprise recorded sound of tapping in the abdominal area of volunteers with a neurosurgical hammer and processed each said recorded signal through the Preprocessing unit and Feature extraction unit;

said random forest classifier based on extracted features of the percussion sound signals of the volunteers enabling identifying the state of the abdomen for the subject.

According to yet another aspect in the present invention there is provided a kit for percussion sound analysis comprising:

A) a percussion sound analysis system for detection of various abdominal conditions comprising (i) a sound acquisition system to capture the percussion sound signals from abdominal area of a subject; (ii) a computing unit to process and analyze the percussion sound signals including preprocessor of signals, feature extractor and classifier enabling detection of abdominal conditions in the subject; and (iii) a cooperative user interface; and
B) a neurosurgical hammer.

BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS:
Figure 1 shows a functional block diagram illustrating basic operational steps of percussion sound analysis system in accordance with the present invention.

Figure 2 shows block representation of overall acquisition system of the present invention.

Figure 3 shows block diagram of training model development process in accordance with the present invention.

Figure 4 shows block diagram showing process for classifying new samples in accordance with the present invention.

Figure 5 shows Visual Analysis Window in accordance with the present invention.

Figure 6 shows Flowchart of the Percussion Sound Analysis in accordance with the present invention.
Figure 7 shows screenshots of Different Parts of the Invented System.

Figure 8 shows receiver operating characteristic curve for the trained classifier in accordance with the present invention.

DETAILED DESCRIPTION OF THE INVENTION WITH REFERENCE TO THE ACCOMPANYING DRAWINGS:
As discussed herein before, the present invention discloses a non-invasive small and cheap Percussion Sound Analysis system or percussion sound analyzer to identify various abdominal ailments that a subject may suffer from. The analysis system includes two operational phase. First, enrolment of the subject i.e. human or patient or user of the present system, wherein a trained classifier is created with the percussion samples recorded. Second, the classification stage, where percussion signals are recorded from the subject and then analyzed to identify the state of the abdomen of the subject.

Reference is first invited from the accompanying figure 1 which shows a functional block diagram showing basic operational steps of the present percussion sound analyzer (100) for abdominal state detection.
As shown in the accompanying figure 1, the Percussion sound signals are captured by electronic means such as sound accusation unit directly from the subject (107). The sound accusation unit captured Percussion sound signals are then processed and analyzed in a computing unit by executing preprocessing (102) the signals, extracting features of the signals (103) and classifying the features (104) based on trained classifier or model (106) in order detect (105) abdominal ailments in the subject. The recorded signal is visualized via waveform or via extracted and analyzed features using a GUI.
Reference is now invited from the accompanying figure 2, which shows a block representation of overall acquisition system (200) of the present invention.
As shown in the accompanying figure 2, the Percussion sound signals capturing means basically includes acoustics medical device (201). The acoustics medical device (201) pickup the Percussion soundsignals from the body and transmit it to the signal conditioner unit (202) for pre-emphasizing the signals.
The signal conditioner unit (202) consists of pre-amplifier, filter and instrumentation amplifier. In the Pre-amplifier, the captured Percussion sound signals at 0-2 kHz frequencies are amplified. The subsequent filter further enhance the Percussion sound signals with frequencies in the range of 0-2 kHz and at the same time suppress the high frequency (above 2 kHz) noise components. The instrumentation amplifier eliminates the need for impedance matching and thus made the unit particularly suitable for use in measurement and test equipment.
The signal conditioner unit (202) feeds the conditioned signals to the A/D converter (203) and the subsequent computing unit having a Digital computing device (204) and a Digital signal processing unit (205) for analyzing the signals. The Digital signal processing unit (205) is disposed in operative communication with an auditory unit (206) and a visual display (207) for showing the analysis result.

Training Model Development:
Development of training model is an important stage in the operation of the present system. During the development of the Training Model, volunteers are tapped over the abdominal area of the subject with a neurosurgical hammer and the signals are collected and recorded with stethoscope stuck to the skin by means of a micro-tape. The stethoscope is placed adjacent to the percussion spot. The tapping is applied through a sheet of plywood around 2.5 mm in thickness (plessimeter) pressed against the skin.
Each recorded signal is preprocessed, followed by segmentation and feature extraction. The generated feature set is used to train a random forest classifier. This trained classifier is later used to identify the state of the abdomen for a new subject. The process flow for developing training model is shown in the accompanying figure 3.

Unknown Sample Classification:
For every new subject, firstly, details like name, age, height and weight are taken. Then, the signal is recorded, followed by preprocessing and feature extraction. Then, with the help of previously trained random forest classifier and WEKA open source libraries, the signal is classified as ‘empty’ or ‘full stomach’ based on the number of segments favoring each of those classes. The process for classifying new samples is shown in the accompanying figure 4.

In the present invention, the computing unit executes preprocessing (102) of the signals in four stages during analysis of the Percussion sound signals.
In the first stage, each percussion blow is segmented from the collected signals in order to study its characteristics. The segmentation is carried out in a segmentation unit by first finding peaks of the signal to identify the regions of percussion blow. Then, from the position of each peak, a marker is moved sufficiently to left side to ensure that the start of the signal is not missed. Now, when the marker starts moving to the right, slope of the signal is measured and onset of the percussion signal is identified by using a threshold slope value. The threshold slope value is fixed after observing the onset of numerous percussion blow signals. Once the onset is identified, a fixed length of the signal ( 10,000 samples ~ 225ms ) starting from the onset point is segmented. Signals belonging to empty stomach state were labeled ‘0’ while full stomach states were labeled ‘1’..
In the second stage, the segmented signals are decimated by using a sampler. Since only the frequencies up to 1kHz are of interest, the sampling rate of the signal could be reduced, leading to lesser number of samples and lesser processing times. Thus, the segmented signals are decimated by a factor of 20 to reduce their sampling frequency to 2205 Hz and their length to 500.
The third stage is amplitude Normalization. Since the percussion blows are generated by tapping the abdomen with hammer manually, the intensity of different blows varied. In order to nullify this effect, each segment is normalized by using a normalization unit to have maximum amplitude of ±1.
In fourth stage, the noise is filtered. In noise filtering, the signals are passed through a 2nd order Butterworth band pass filter with cutoff frequencies 20 Hz and 1 kHz in order to eliminate the noise.

After completing the preprocessing, the computing unit extracts features from the segmented signals. Each segment of the signal contain exactly one percussion blow which is used to extract certain general features, such as Maximum peak to peak amplitude, Peak Frequency in the segment, Total power in each segment, Mid-power frequency (Frequency below which half of the total power is contained).
Normally, the segmented signal has exponentially decaying nature. With this nature of the segmented signal, each segment is approximated by a sum of decaying sinusoids. Observing the frequency spectrum of the signal, it can be seen that there lie various frequency peaks in the frequency range of interest (20 Hz to 1 kHz). To obtain these peak frequencies, the signal is approximated as an exponentially decaying sinusoid after dividing it into overlapping frequency sub-bands by filtering it through the following filters:
A 2nd order band pass filter with cutoff frequencies 10 Hz and 230 Hz.
A 2nd order band pass filter with cutoff frequencies 220 Hz and 420 Hz.
A 2nd order band pass filter with cutoff frequencies 410 Hz and 560 Hz.
A 2nd order band pass filter with cutoff frequencies 550 Hz and 1000 Hz.
The major frequency components in each of these filtered signals are used as features. This gave rise to four more features. In order to account for the varying thickness of skin tissues across individuals, the BMI of individuals is also taken as feature.
Subsequent to the extraction of the features, the computing unit involves Random Forest Classifier for further analyzing the Percussion sound signals. The random forest starts with a standard decision tree. In a decision tree, an input is entered at the top and as it traverses down the tree the data gets bucketed into smaller and smaller sets. The random forest takes this notion to the next level by combining trees with the notion of an ensemble. The procedure for training of such a system, for some number of trees T, is as follows:
2/3rd of the training cases are sampled at random with replacement from the training set to create a subset of the data
At each node, for some number m, m feature variables are selected at random from all the feature variables. The feature variable that provides the best split, according to some objective function, is used to do a binary split on that node.
At the next node, another m variable is chosen at random from all predictor variables and same operation is performed.

When a new input is entered into the system, it is run down all of the trees. The result is a voting majority (when decision threshold is set as 0.5) of all of the terminal nodes that are reached.
In the present work, initially percussion sound samples were collected from various subjects in two different states- ‘Empty Stomach’ (class 1) and ‘Full Stomach (450ml of water)’ (class 2), and an attempt was made to identify the states based on the signals. From the training samples, 2397 segments are obtained after segmentation. Features extracted from these segments are used to train the random forest classifier. In order to test the classifier, a 10-fold cross validation method with features extracted from individual segments of training data is performed. The results obtained are as follows:

Accuracy Sensitivity Specificity AUC
90.6133 % 88.9 % 92.2 % .967

Visual Analysis:
Instead of using the previously trained classifier, the user has the option of making use of visual analysis to make a manual decision on the state of the abdomen. The visual analysis window as shown in the figure 5 illustrates the comparison of the feature metrics as analyzed by the computing unit with those observed in past for both states of abdomen using a parallel co-ordinates graph.
The accompanying figure 6 shows Flowchart of the Percussion Sound Analysis Tool and the figure 7 shows Screenshots of Different Parts of the Invented System.
The block Screen-1 is the first window of the GUI developed. Block Screen-2 is used to input the details of a new subject before proceeding to percussion analysis for him/her. Block Screen-3 is used to record the percussion signal of a subject. Block Screen-4 can used to observe the Fourier transform of the recorded signal, listen to pre-recorded sample sounds for empty and full stomach and to proceed for further analysis.

The present system has been tested with 7 subjects out of which 3 were ‘empty stomach’ and 4 were ‘full stomach’. All subjects were males. In 6 out of 7 cases, the output results were correct while 1 sample was misclassified.

Subject No. Original State Determined State by Model Confidence Level Classification Result
1 ‘Empty Stomach’ ‘Empty Stomach’ 100% Correct
2 ‘Empty Stomach’ ‘Empty Stomach’ 100% Correct
3 ‘Empty Stomach’ ‘Empty Stomach’ 94.73% Correct
4 ‘Full Stomach’ ‘Empty Stomach’ 88.88% Wrong
5 ‘Full Stomach’ ‘Full Stomach’ 91.95% Correct
6 ‘Full Stomach’ ‘Full Stomach’ 74.07% Correct
7 ‘Full Stomach’ ‘Full Stomach’ 80.76% Correct

These results thus clearly indicates that the system of the present invention execute the percussion sound analysis technique to detect various abdominal conditions in a subject in much superior manner compared to the existing percussion sound analysis techniques. The, advantageous aspects of the present system can be summarized as hereunder:
(i) Does not rely on the experience of the doctor and is able to overcome the natural limitations of the human ear.
(ii) Trained technician is not required; layman can handle this system.

(iii) The system is non-invasive, cheap and small in size.

(iv) Results are verified using two ways: first one is machine learning based suggestion and second one is visual analysis.
(v) A permanent record of the observation is present as opposed to age-old method where the doctor had to make spontaneous decisions based on what he hears.
(vi) The system has a potential to acquire a large amount of information about the both physical nature and functionality of the repository system which could be exploited in future.

It is thus possible by way of the present invention to provide for a superior and technically advanced percussion sound analysis system which can be used to analyze percussion sound captured from abdominal region of a subject in order determine varying ranges of abdominal aliments in the subject. Additional modifications and improvements of the present invention may also be apparent to those skilled in the art. Thus, the particulars combination of parts described and illustrated herein is instead to represent preferred embodiments of the present invention, and is not intended to serve as limitations of alternative devices or/and combinations within the spirit and scope of the invention.