Recent Trends in Image Processing and Pattern Recognition

WHO projects that diabetes will be the 7th major cause leading death in 2030. Diabetic Retinopathy caused by leakage of blood or fluid from the retinal blood vessels and it will damage the retina. For detection and extraction of non-proliferative diabetic retinopathy lesion we have invent the new wavelet filter. The proposed filter give the good extraction result as compare to exiting wavelet filter. In proposed algorithm, we have extract the microaneurysms, hemorrhages, exudates and retinal blood vessels. After extraction of lesions, grading is done by using feed forward neural network. The proposed algorithm achieves sensitivity of 98%, specificity of 92% and accuracy of 98%.

Skin lesion is the most deadly skin disease in humans, it arises as a result of disorders in the pigment cells, which is produced by a pigment known as melanin. This disease can be prevented and treated if there is an early diagnosis of the disease. Computer-Aided Diagnosis (CAD) has played a key role in helping dermatologists to diagnose the disease. In this proposed system, the model for diagnosis and classification of lesions consists of several stages beginning with pre-processing for the purpose of enhancing images, and identify the area of the lesion by isolating it from the healthy body, and extract features from an region of interest using the LBP method. Classification of the lesion into any of the three classes belong, benign or atypical or malignant according to database PH2. The results obtained for both SVM and K-NN classification techniques were 93.42% and 96.05% respectively.

Magnetic resonance image is a greatly developed medical representation system, cynical to generate prominent renowned imagery of human being and there part. It gives feature in sequence to analyze the disease. MRI play significant role to give in turn innovative scope intended such precede representation. Inventive image of MRI are usually comprise squat disparity. It is dense intended in favor of health center to analyze them. Escalating the difference of representation, it resolve exist simple designed in favor to analyze a complete in sequence. This manuscript compares special method of development of brain MRI by histogram based technique and by using different statistical measures.

Osteoarthritis (OA) is one of the chronic diseases related to joints. The joint pain caused due to Osteoarthritis is unbearable and if not treated may cause deformity and disability. It is observed that few numbers of researchers have implemented identification and grading of Osteoarthritis utilizing diverse methodologies based on their own datasets for experimentation. However, there is still need of automatic computer aided methods to detect Osteoarthritis for early recognition. In this work, cartilage region is automatically identified based on density and range of wavelet filters were used for appropriate and early analyses of radiographic Knee Osteoarthritis images. The computed wavelet features are classified using decision tree and k-nearest neighbour classifiers. The indicated experimental outcomes are challenging and viable which are examined by medical experts.

Nowadays not only older people but also younger one (having age in between 30 to 45) are also suffering from knee Osteoarthritis (OA). Diagnosis and proper treatment of OA can delay the onset of severe disability. The primary pathological features (reduction in joint space distance and synovial cavity, inflammation and change in bone morphology for instance) to predict knee OA severity is clearly visible in X-ray images. However the X-ray images findings in this disease are often non-productive and may mislead due to the presence of noise in the image and create difficulties for analysis and classification. So to overcome the aforementioned problems an effective OA classification is system is presented.The proposed classification system is to work in stages. Initially X-ray image is enhanced by using contrast stretching-based intensity transformation function. Next, the histogram modeling-based segmentation method is adapted to localize the desired region that is the knee joint region. Extract ROI is converted to discrete wavelet transform to do multi-resolution analysis which helps to detect interesting patterns those are not visible in an input image. As the inflammation cause the change in texture and morphology in affected knee joint region, DWT based statistical features which represent textural changes are extracted and passed to the SVM model for exact severity grade detection. Experiments are conducted on patient-specific X-ray images procured from local sources (Chidgupkar hospital Pvt. Ltd. India). In total hundred images are collected of various severity grades. By confirming results our model yields 99% accuracy for five-class classification and it will be a beneficial tool for rheumatologists to predict OA severity grade. In future we aim to conduct experiments on Osteoarthritis database (OAI) to confirm the robustness of the proposed system.

Criminal activities have increased largely over the past couple of years and the security of the commoners especially women have been hugely jeopardized. There has been multifarious cases of threats and assaults with weapons in the present days especially with knives which are one of the most common and readily available household items. Such cases have made CCTV cameras a common sighting in the neighbourhood. The prime idea behind their installation is surveillance. The footage from such cameras can serve as an extremely important source of evidence during investigation. However, such systems only make themselves useful as evidences of a crime and do not aid in prevention of a crime in progress. Standing in such times, making CCTV cameras intelligent can be a solution which can detect weapons and thereafter alert authorities. Here, a deep learning based system is presented which can automatically detect visible knives to alert authorities of a prospective crime and thereby aid in women security. The system has been tested on a freely available dataset [5] consisting of over 12000 frames and a highest accuracy of 96.11% has been obtained. We have also tested the performance of handcrafted feature-based framework with grey level co-occurrence matrix (GLCM) and our system produced better result.

For the medical domain, computer assumes a significant job in computerization and determination of the disorder. The stethoscope is an eminent and widely available traditional diagnostic instrument for the medical professionals. The computer system is used in medical science for collection and analysis of large amounts of massive data and concern accurate decision making. The respiratory sound database has been available from research community. However, full utilization of available recording device or database, there is a need to design and development of the respiratory disease identification. This paper explained the respiratory data creation and application of this data over the respiratory disorder identification. The database is collects with the help of local government hospital. The data is recorded with directional stethoscope with 3.5 jack based microphone connected with laptop or computer. The database includes 1000 recording of 7.5 h. The data is collected from 50 patients. The Mel Frequency Cepstral Coefficient technique is applied over the database for feature extraction. The pitch, energy and time are the dominant features for the disorder identification. The neural network has been used for the classification of the disorder identification. The experiment has been achieved accuracy of 91% over the two class classification. The precision of the experiment is 88% whereas sensitivity is 87%. The 9% error rate has been shows the experimental system. From the experimental analysis the author recommended the MFCC and neural network are the strong and dynamic approach in respiratory dieses determination.

The Advancement of communication system has given us the freedom to think beyond traditional communication system and stage is set for thought oriented communication system. There are thousands of thoughts generated and vanished in a timeframe but out of these some prominent thoughts persist and we proceed with the same in our day to day activities. The advancement in Electroencephalogram has provided a chance to see the activity in the human brain in non-invasive manner. The proposed research work presents the method for Digit recognition using the EEG signals acquired and processed on smart devices. The results show the implementation of Computation neural network for the recognition of digits from EEG signals. It was seen that, the 90.64% correct classification was achieved.

Cardiac arrhythmia is very harmful heart disease which can cause severe and even potentially deadly symptoms. Early diagnosis of an arrhythmia can save lives. In the modern health care environment, the use of Internet of Things (IoT) technologies brings the suitability for medical professionals and patients, because they are useful in various medical background. An arrhythmia diagnosis system based on IoT sensors helps and monitors automatic transmission and measurement of Electrocardiogram (ECG) signal data, analyzes this data, and alerts medical professional for an urgency. We are developing a new approach for monitoring and diagnosing cardiac arrhythmia using IoT environment. The concerned ECG signal data is generated by using the IoT sensors and standard online (UC Irvine repository) dataset for predicting severity of cardiac arrhythmia. Also, we used a new fuzzy logic-based neural network classifier algorithm for diagnosing cardiac arrhythmia. The experiments are carried out by various machine learning algorithms on standard UCI Repository dataset as well as on real patients data records collected from multiple Indian hospitals. The proposed machine learning algorithm offers the best accuracy than different machine learning approaches with minimum time complexity. Patients can check their heart condition at their home by using this system. The maintenance and development cost of this system is low due to it is a lightweight and small size system.

For the heart diseases, the early prediction required to save the human being life. There are several ways to perform the early prediction of Cardiovascular Disease (CVD), however the most of the state-of-art approaches are expensive with poor accuracy of prediction. The computerised approach used the Electrocardiogram (ECG) signals to perform the early prediction of CVD. The ECG based approach is simple, effective and inexpensive; hence it gains the significant attention of researchers from last two decades. The Computer Aided Diagnosis (CAD) system introduced the ECG based approach for CVD prediction using the ECG signal of patients on which the algorithms single processing, data mining, and machine learning applied for accurate prediction. ECG based CVD detection framework composed of three main sections that is preprocessing, features extraction, and classification. The steps like preprocessing and features extractions are crucial for the efficiency of CVD detection. In this paper, we proposed the novel framework of CVD detection of Q, R, S, T beats efficiently from the pre-processed ECG signal. From the pre-processed ECG signal, our aim is to extract QRS and ST segments using the dynamic and simple thresholding approach. The segments are used then for the statistical features extraction. The classification is performed by using the Artificial Neural Network (ANN) classifier. The proposed method shows the precision rate, recall rate, detection accuracy and detection time are 0.91, 0.92, 0.91 and 1.51 respectively. It shows the balance between the accuracy and prediction time performance as compared to state of art method.

Lung cancer is among the dominant cause of deaths due to cancer. Lung cancer survival largely depends on the stage at which it is diagnosed with early stage diagnosis significantly improves the survival rate of patients. Radiologist diagnoses the Computerized Tomography images by detecting lung nodules from the images. Detection of initial stage lung cancer is very challenging as the sizes of lung nodules are very small and are difficult to locate. Many computer aided detection systems to detect lung nodules were proposed to assist radiologist. Recently, Deep learning neural network has found its way into lung nodule detection system after the success it has exhibited in computer vision tasks. In this paper, we propose a novel deep convolutional neural network based system for lung nodule detection and localization. Our objective is to provide radiologist with a tool to correctly diagnosis Computer Tomography (CT) scans of patients. The system developed was able to detect and localize with the sensitivity of 92.9%. LIDC-IDRI world largest publicly available database for computer tomography scans of human lungs was used for this research.

In this work, fuzzy method is proposed to enhance the contrast of Magnetic Resonance Imaging (MRI) brain images. Negative Image (NI), Log Transform (LT), Gamma Correction (GC), Histogram Equalization (HE), Adaptive Histogram Equalization (AHE) and Dynamic Histogram Equalization (DHE) methods are compared with proposed method. The performance is evaluated by using quantitative measures like Michelon Contrast (MC), Entropy, Peak Signal to Noise Ratio (PSNR), Structure Similarity Index Measurement (SSIM) and Absolute Mean Brightness Error (AMBE) as a parameter on BRATS-2014 dataset. The proposed method gives good results for Entropy, PSNR and AMBE, we need to improve the proposed method for MC and SSIM.

Radon transform is one of the features used in image reconstruction from the early days of image processing, later it has been used in many applications of astronomy and other fields, such feature is explored for the process of matching and retrieval of images. The proposed system has 4 phases, pre-processing, feature extraction, knowledgebase construction and image retrieval. The local features such as sum, mean, standard deviation, & the eigen values are extracted for each radon transformed image and stored in a knowledgebase. Cityblock distance measure and query-by-image method are employed for matching and retrieval of images respectively and the accuracy of 87% is achieved.

In this paper, we propose a new contrast optimal visual Cryptography Scheme for Half-tone images. This scheme can share Secrete Half-tone Image among n members and all n number of members can reveal the secrete by superimposing the n share together and n − 1 members can not reveal the secrete. The proposed method generate the noise like shares but share size and the image recovered after superimposition is of same size which is equal to original secrete image. It reduces the pixel expansion based on the designed codebook and matrix transposition. It also reduces the transmission speed, complexity computation, security issue and restore the secrete image without any distortion. The scheme proposed can share the black& white and gray level image and produces the more accurate results and is easy to implement.

The main focus of the presented work was to better predict the surface mineralogy from the Chandrayaan-1 hyperspectral data set covering the area from South Pole region. To address the space weathering effect and to quantify mineralogy the Bi-directional reflectance function have been implemented. The implemented model was tested against two standard lunar laboratory mixtures and with the Apollo 10084 bulk soil sample. About 85 spectra were initially selected from varying locations and only active spectra with significant absorption were used for modeling. The minerals like plagioclase and Clinopyroxene were identified. Many spectra exhibits more iron content simulating mature area. Model result show no olivine content and very low Orthopyroxene content may be because of more crustal thickness, no impact would have penetrated to the lower mantle. Study reveals the potential of hyperspectral data multiplexed with mathematical model for not only mineral quantification but also helps to predict other associated parameters like grain size, iron fraction, phase function, however the spectra from mature soil and the limited HySI coverage acts as challenge for modelling process, modeling the data at longer wavelengths will be an advantage to improve the accurate mineral prediction.

The microwave Synthetic Aperture Radar (SAR) is an active type of remote sensing. The classification analysis has become one of the very important task, after the availability of microwave SAR datasets from the satellite. The one of the major challenges faced is the accuracy regarding classification analysis. In the present paper the two supervised classification techniques used, i.e., Wishart and Support Vector Machine (SVM). The accuracy results for both classifiers are analyzed on the basis of confusion matrix and omission, commission error. The overall process is carried out on SIR-C L-band SAR dataset of Kolkata (W.B.) India. The four major classes studied is Water, Trees and Mangrove, Paddy, Settlement. The present work focus on the agricultural application. From the overall work it is found that the accuracy of the Wishart supervised classifier is 92.18% and for SVM supervised classifier it is 99.58%. There is very huge difference of 07.40% between these two classifiers. Hence the SVM classifier has better accuracy compare to Wishart classifier. The overall work done by using software tool PolSARPro Ver. 5.0 and NEST Ver. 5.0.16.

The planet earth is house of a variety of species of bird. Many of these birds are now getting extinct. Moreover, a lot of wildlife crimes are committed against birds including shooting, trapping, poisoning, and illegal sale of rare species. Feathers may be good evidence in such cases to identify species of birds. In this study, we propose a pattern recognition based technique for identification of birds from feathers. Our literature survey could not reveal a systematic study on identification of birds from the images of feathers. We have made a digital database of 60 feathers from 15 different species of birds. Hue and saturation histogram, which yields a feature vector of 46 dimensions, is extracted from the images of feathers. Nearest Neighbor (NN) algorithm is utilized for identification of birds using various distance metrics, which resulted into a maximum accuracy of 95.46%.

The paper is intended to transform the voice-signal from the frequency domain into a spatial domain in form of grayscale image and applied the image processing techniques. To satisfy our hypothesis, two models of signal processing were carried out in this research: Speaker Recognition and Signal Segmentation. For applying the image processing techniques on the voice-signal, two methodologies were proposed to convert the signal into grayscale-image: signal-range based and fuzzy-based. The signal-range based is to convert the signal range from (−1 ↔ 1) into (0 ↔ 256). The second method of conversion, Fuzzy Gaussian Membership Function is applied to convert the signal range into (0 ↔ 1), then multiply them by 255 to be in the range of grayscale image. In the Speaker Recognition, the LBP is used as pre-processing for filtering the intensity of the signal image. The HOG is used to extract the features of signal-image. So, the total length of features-vector is 324. The classification learner tool in MATLAB was used for classifying the feature-vectors and the results were found to be satisfactory. The automatic word segmentation was proposed based on thresholding and morphology operators. The segmentation accuracy is 93.67% in the Marathi-language. The highest recognition rate in speaker identification system is 96.9%.

Process of planting crop is the main source of revenue of rural India as most of the population depends on farming for means of support. but due to the annoying characteristics of diseases, and unforeseen changes in the climatic conditions, the overall yield of the produce reduces. At large this will avoid losses both in terms of value and volume of the agricultural harvest. The Regular methods of investigating the infection in yield are not efficient and are not precise enough in detecting the infection associated with plants. An expert system that can correctly and effectively identify and detect plant infection will helps to enhance the yield quality and quantity of plants

Geocell is being used very conveniently for soil reinforcement and slope protection in the recent times. The main advantage of this synthetic material is provision of the confinement to the soil resulting in the higher resistance to shear failure of the soil. Geocell is a three-dimensional synthetic material having the varying wall thickness and opening sizes. The use of the confining properties of the geocell is needed to be investigated in the application of bearing capacity improvement of soil below the foundations. The present work demonstrates bearing capacity improvement of the soil using geocell as the confining reinforcement. A series of experimental model tests is performed in the present study with constant soil type (sand). Parameters considered for variation was the width (W), opening size (og) and height (hg) of the geocell. The image analysis was performed mainly in ImageJ which is available on free license program with Digital Image Correlation and the Particle Image Velocimetry approach. The observations were made mainly in terms of maximum load carried by the subsoil and the depth of Rankine’s zone obtained due to placement of geocell. Load bearing resistance of subsoil was in proportion with the width and height and in inverse proportion with the opening size of the geocell. The Rankine’s zone also observed to shift downward due to geocell inclusion providing more shear surface and thus higher bearing capacity. Thus, the geocell may successfully be utilized below shallow foundations to obtain higher Safe Bearing Capacities.

Plant diseases have grown-up in agriculture to be an impasse as it can cause dwindling in both quantity and quality of farming yield. This work explains a simple and adept method used to recognize leaf diseases by applying digital image processing and machine learning technology. In this study 24 color feature, 12 shape and 4 texture features were obtained from images of four kinds of diseases like, Sigatoka, Panama Wilt, Bunchy Top and Banana Streak Virus diseases. Principal component analysis (PCA) was achieved for reducing dimensions in features extracted and k-nearest neighbour classifiers used to identify banana diseases. The finest result was obtained when image identification was conducted based on PCA with K-nearest neighbour classifier.

Plant species database has become essential as biodiversity is declining rapidly. With advance technology and technocrats, an attempt has been made of plant species identification depending on leaf image is implemented. Leaf characteristics are used to prepare feature vector. HOG and LBP are used as feature vectors, LDA and SVM as classifiers. When HOG and SVM are concatenated and used as feature vector, accuracy of classifiers is enhanced as compared to individual HOG and SVM as feature vector. LDA is proved to be better classifier than SVM for database mentioned.

Scare water resources necessitates technological involvement in irrigation scheduling, that can help to manage water according to the weather condition of different seasons, crop growth stage and landscape information. The proposed method calculates actual water required using machine learning model and Evapotranspiration. The model is trained using real time weather data to predict actual water requirement. Reference Evapotranspiration is calculated with the help of Penman-Monteith Method. Before starting with this real time system proposed model is implemented with the help of past 10 years web scraped weather data. Proposed algorithms of water requirement and Irrigation scheduling is executed on scrapped data. After successful results system is implemented for real time use.Furthermore, system consists of eight Arduino nodes that acting as a slave to read weather, soil landscape and rain data. In addition, three Raspberry-pi equipped with the Wi-Fi module, acting as server to send collected data to a remote web server. These databases are used as input for machine learning algorithm. As per the observations of proposed system water usage is getting reduced in large quantity as compared to the traditional irrigation system used for irrigation.

In this proposed work we have focuses on the estimation of soil moisture using L-band and C-band polirimetric (HH, HV, VH, VV) and dual band (VV, VH) SAR (synthetic aperture radar) data set. The empirical model derived by Oh for finding scattering parameter from bare soil surface was implemented. Polarimetric radar backscattering were conducted for bare soil surface for a different values of surface roughness and surface soil moisture at L-band and C-band at different incidence angle from 20° to 60°. A series of field data were collected from different areas as per the standard method given in literature. The data in this paper were taken from two days near Bardoli and Ahmedabad city during the field experiment conducted in 2017. Currently we analyze the collected data to understand the relation between field parameter (surface soil moisture, dielectric constant, surface roughness) and SAR data (radar backscattering). In addition radiative transfer model and radar backscattering model are used to simulate the L-band, C-band data observations. In this work we have simulated empirical Oh model for estimating surface soil moisture using L-band, C-band SAR data set. We have applied all the models given by Oh in different research paper between (1992–2002). Further this results were used to find dielectric constant and use Topp’s model to derived surface soil moisture. A good agreement was observed between the estimated and simulated values. Performance of Oh model is promising for L-band and quite good for C-band SAR data for surface soil moisture estimation.

In present work, there are three wireless sensor nodes developed for the measurement of microclimate at three different locations inside greenhouse by designing a wireless sensing node that can measure various atmospheric parametric conditions like carbon dioxide, Oxygen, Humidity, Temperature and light intensity inside greenhouse as well as Soil moisture contents. The designed and developed nodes were placed in a star topology inside greenhouse at Sangvi village at Baramati tehsil and district Pune in Maharashtra state of India. During experimentation, data was collected at different times and on various dates for proper testing and evaluation.

The aim of this research is to highlight the advantages of Climate-Smart Agriculture and the progress accomplished by implementing information technology to make agriculture intelligent. The research article also includes the circumstances for policy and investment to succeed under climate change in sustainable agricultural growth for food security. It also involves a cropping calendar that differs for males, females, and kids to classify the gender division of labor and access and resource control. Climate information is taken from various tools such as the Department of Meteorology, Satellite Images, WSN, and IoT Tool Kit. Accordingly, tree cultivating, preservation agriculture, minimum tillage, and natural resource management are accumulated under the single umbrella of Climate-Smart Agriculture (CSA). The aim of this paper is to have a glance of relationships between CSA and its application.

This paper describes the importance of machine learning (ML) algorithms in the design and development of diagnostic expert systems. The designed system is intended on how ML techniques can be used in the prediction of pig diseases through visible symptoms and their past behaviours. These types of systems are very useful in situations where the domain experts in the field are not readily available. It is important to make an accurate diagnosis of pig diseases to provide control strategies for any symptom shows in a pig to prevent related health issues. The main concerned in the study begins with the collection of disease symptoms and related information of pigs during their life span from the domain experts and literatures in the domain. The acquired knowledge and information are then pre-processed to be used for being trained different ML algorithms. The system development involves the training and analysis of different ML classifier models and their performances. It has thus been found in the study that SVM classifier can produce more accurate results in disease prediction. The paper finally presents the design and development of an expert system for the diagnosis of Pig diseases using SVM classifier which is trained with pig disease dataset.

Modern plant identification system highly depends upon robust features and classification algorithms possibly the combination of several modalities. In this paper, we have used deep CNN features as well as we introduce venation angles present between two veins as a features to classify plants. We have employed 5 types of classifiers to classification and results have been compared using evaluation measure Mean Reciprocal Rank (MRR) and F-measure. VISLeaf dataset have been used to perform the experiments, the system achieved the accuracy of 96.67%, 92.76%, 80.00%, 53.89%, 96.67% for SVM, KNN and Naïve Bayes, Tree classifier, Neural Networks, respectively and 97.22% for proposed hybrid votes classifier.

Plant leaf disease detection & classification is a complex image processing task, wherein proper algorithms are needed for segmentation, pre-processing, feature extraction and classification. Generally linear classification algorithms like support vector machines. (SVMs), k-nearest neighbour (kNN), Naïve Bayes (NB), Random Forest (RF), etc. do not provide high precision for classification when applied to leaf disease classification. This is due to the fact, that the features which are evaluated during the segmentation and feature extraction phases are do not vary much in terms of values, but they vary in terms of patterns of occurrence. For example, leaf images which are taken for bacterial blight and Alterneria do not show significant changes in feature values, but they show major changes in feature patterns, which is generally neglected by these linear classifiers, and thus the accuracy reduces. In order to improve the accuracy of classification, we propose a hybrid convolutional neural network (CNN) in this paper, which combines multiple methods of segmentation & feature extraction with CNN in order to improve the accuracy of the system. The developed system shows 22% higher accuracy than the existing systems, and can adapt to any type of leaf images by moderate level of training.

In the prevailing period of innovation the automatic identification of speaker assumes a significant role. The application of speaker recognition is spin towards Biometric security. This paper depicted a speaker recognition framework for isolated word dataset. The feature extraction has been finished utilizing Mel Frequency Cepstral Coefficient (MFCC) techniques. The database of the research is structure and creates utilizing 25 Male and 25 Female speakers. The size of dataset is 2500 isolated words. The content for the dataset recording is chosen based on vowel letters in order. The execution of the framework is determined utilizing False Rejection Rate (FRR), False Acceptance Rate (FAR). The precision of the Speaker Recognition rate for Male is better as compare with the exactness of female. This structure is utilized for speaker recognition framework for the confined word distinguishing proof framework by applying highlight extraction methods as MFCC and arrangement is finished with Euclidian Distance. We got a normal exactness for Male rate is 85% and 81% for female. The exhibition of the haphazardly chosen subject gathering was 79%. This is the general precision pace of Speaker Recognition framework for Marathi Isolated Words.

Speech recognition is to convert speech signal into text. It is challenging task due to natural variations present in human speech and also due to background noise. Now a day’s researchers focus on deep learning due to it’s effectiveness and high performance. There is need to work on recognition of speech for mathematical words; many of the currently available methods proposed by researchers are not yet adequate to recognize speech for mathematical words. In this paper we focus on CNN (Convolution Neural Network) for recognition of connected word speech recognition for mathematical words plus, minus, square and square-root. The dataset of spoken words are created using Audacity. CNN model is verified by considering Adam, Gradient descent and Adagrad optimizer with learning rate of 0.001 and 0.0001. Result shows that Adam and Gradient descent optimizer gives better result for 0.001 learning rate.

Notation of a music piece is an extremely important resource for musicians. It requires mastery and experience to transcribe a piece accurately which lays the path for automatic music transcription systems to help budding musicians. A piece can be divided into two parts namely the background music (BGM) and lead melody. The BGM is an extremely important aspect of a piece. It is responsible for setting the mood of a composition and at the same time makes it complete. There are different musical instruments which are used in a composition both in the BGM and lead sections one of them being the bass guitar. It bonds with the percussion instruments to form the spinal cord of a piece. It is very much important to transcribe the bass section of a composition for understanding as well as performance. Prior to identification of the notes being played, it is essential to distinguish the different patterns/grooves. In this paper, a system is presented to differentiate bass grooves. Tests were carried out with 60K clips and a best accuracy of 98.46% was obtained.