CLIAMS:I/We claim: 1. A method for monitoring a physiological parameter associated with a subject using a hand held device (130), the method comprising: extracting, by a processor (102), a plurality of PPG features from a video of a body part (128) of the sample subject, the plurality of PPG features being associated with the physiological parameter; determining, by the processor (102), a correlation coefficient for each of the plurality of PPG features, indicative of a relation between a PPG feature and a ground truth value of the physiological parameter; ascertaining, by the processor (102), a gain factor for each of the plurality of PPG features, based on the correlation coefficient; and selecting, by the processor (102), relevant PPG features from among the plurality of PPG features, based on the gain factor, wherein the relevant PPG features are deployed for monitoring the physiological parameter in real time. 2. The method as claimed in claim 1, wherein the extracting comprises obtaining the plurality of PPG features from the video in at least one of a time domain and a frequency domain. 3. The method as claimed in claim 1, wherein the physiological parameter comprises at least one of a blood pressure, an electrocardiograph (ECG) indicative of heart condition, blood oxygen level, and a respiration rate. 4. The method as claimed in claim 1, wherein the correlation coefficient is a maximum information coefficient (MIC). 5. The method as claimed in claim 1, wherein the ascertaining the gain factor is based on a sigmoid gain function. 6. The method as claimed in claim 1, wherein the ascertaining the gain factor comprises tuning a slope constant (m) associated with the gain factor, based on accuracy of a k-fold validation technique, the tuning being performed using one of a regression model and a classifier model. 7. The method as claimed in claim 6, wherein the regression models is one of a linear regression model, a non-linear regression model, and a polynomial regression model. 8. The method as claimed in claim 6, wherein the classifier models is one of a support vector machine (SVM)-based model and an adaptive neural network (ANN)-based model. 9. The method as claimed in claim 1, wherein the selecting comprises: multiplying each of the plurality of PPG features with the respective gain factor; and selecting the relevant PPG features from among the plurality of PPG features based on a threshold value of each multiplied PPG feature. 10. The method as claimed in claim 1 further comprising ascertaining, by the processor (102), actual relevance of each of the relevant PPG features based on the respective gain factor. 11. The method as claimed in claim 1, further comprising: obtaining, by the processor (102), test PPG features associated with a test subject from a video of a body part (134) of the test subject; and monitoring, by the processor (102), the physiological parameter for the test subject, based on the test PPG features and the relevant PPG features. 12. A feature selection system (100) for monitoring physiological parameters associated with a subject, the feature selection system (100) comprising: a processor (102); a processing module (110) coupled to the processor (102) to extract a plurality of PPG features from a video of a body part (128) of the sample subject, the plurality of PPG features being associated with the physiological parameter; and a feature selection module (112) coupled to the processor (102) to, determine a correlation coefficient for each of the plurality of PPG features, indicative of a relation between a PPG feature and a ground truth value of the physiological parameter; ascertain a gain factor for each of the plurality of PPG features, based on the correlation coefficient; and select relevant PPG features from among the plurality of PPG features, based on the gain factor, wherein the relevant PPG features are deployed for monitoring the physiological parameter in real time. 13. The feature selection system (100) as claimed in claim 12 further comprising a testing module (114) coupled to the processor (102) to ascertaining actual relevance of each of the relevant PPG features based on the respective gain factor. 14. The feature selection system (100) as claimed in claim 12, wherein the processing module (110) obtains the plurality of PPG features from the video in at least one of a time domain and a frequency domain. 15. The feature selection system (100) as claimed in claim 12, wherein the feature selection module (112): multiplies each of the plurality of PPG features with the respective gain factor; and selects the relevant PPG features from among the plurality of PPG features based on a threshold value of each multiplied PPG feature. 16. The feature selection system (100) as claimed in claim 12, wherein the correlation coefficient is a maximum information coefficient (MIC). 17. The feature selection system (100) as claimed in claim 12, wherein the feature selection module ascertains the gain factor based on a sigmoid gain function. 18. The feature selection system (100) as claimed in claim 12, wherein the feature selection module (112) tunes a slope constant (m) associated with the gain factor, based on accuracy of a k-fold validation technique, the tuning being performed using one of a regression model and a classifier model. 19. A physiological parameter monitoring device (132) for monitoring physiological parameters associated with a subject, the physiological parameter monitoring device (132) comprising: a processor; a monitoring module (138) coupled to the processor to, obtain at least one relevant PPG feature having a correlation with a ground truth value of the physiological parameter to be monitored, wherein the relevant sample PPG features are selected from among a plurality of sample PPG features based on a correlation between a PPG feature and a ground truth value of the physiological parameter, and a gain factor determined based on the correlation; ascertain test PPG features associated with a test subject from a video of a body part (134) of the test subject, the video being captured using a camera (136) of the physiological parameter monitoring device (132); and monitor the physiological parameter for the test subject, based on the test PPG features and the relevant PPG features. 20. A non-transitory computer readable medium having a set of computer readable instructions that, when executed, cause a feature selection system (100) to: extract a plurality of PPG features from a video of a body part (128) of the sample subject, the plurality of PPG features being associated with the physiological parameter; determine a correlation coefficient for each of the plurality of PPG features, indicative of a relation between a PPG feature and a ground truth value of the physiological parameter; ascertain a gain factor for each of the plurality of PPG features, based on the correlation coefficient and a sigmoid gain function; and select relevant PPG features from among the plurality of PPG features, based on the gain factor, wherein the relevant PPG features are deployed for monitoring the physiological parameter in real time. ,TagSPECI:TECHNICAL FIELD [0001] The present subject matter relates, in general, to photoplethysmography and, particularly but not exclusively, to monitoring physiological parameters using a hand held device. BACKGROUND [0002] Monitoring of certain physiological parameters and vital signs of a person, such as respiration rate, pulse rate, pulse oximetry, blood pressure, and heart condition, is typically performed in a clinical setting. Generally, it has been observed that if a person is aware that his or her physiological parameters are being monitored, it may cause the person to become nervous or conscious and the physiological parameters may get altered due to the mental state of the person, and therefore may not reflect the true medical condition of the person. Therefore, several unobtrusive techniques for monitoring the physiological parameters have been developed. [0003] One such unobtrusive technique is photoplethysmography (PPG). PPG involves an optical methodology and works on the basis of dynamics of blood volume changes in the vasculature under the skin. Conventionally, PPG is implemented in various ways for measuring and monitoring physiological parameters, for example, by contactless recording of videos of the subject whose physiological parameters are to be measured. BRIEF DESCRIPTION OF DRAWINGS [0004] 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 components. [0005] Fig. 1 illustrates a physiological parameter monitoring device coupled to a feature identification system for monitoring physiological parameters associated with a subject, in accordance with an implementation of the present subject matter. [0006] Fig. 2 illustrates a method for selecting relevant PPG features for monitoring physiological parameters associated with a subject, in accordance with an implementation of the present subject matter. DETAILED DESCRIPTION [0007] The present subject matter relates to monitoring physiological parameters associated with a subject, using a hand held device. [0008] Conventionally, photoplethysmography (PPG) is implemented in various ways for measuring and monitoring physiological parameters. In few conventional techniques, a PPG waveform is extracted from a video by splitting the video into frames and then a peak frequency of one or more quantized colour values is analyzed in a predetermined set of frames. Generally, an input from a sensing device, such as a pulse oximeter or a sound-based sensor, say for measuring heart rate, is usually used in conjunction with the video for enhancing accuracy of the measurement of the physiological parameters. However, the cost and portability of such an apparatus is adversely affected. In few other conventional techniques, two or more PPG waveforms are extracted from the video and used for the measurement of the physiological parameter. However, in such cases, the processing and analysis of the PPG waveforms uses large amounts of computational resources and time, rendering the technique cumbersome and time consuming. [0009] In order to reduce the computational complexity, other conventional techniques involve extraction of certain PPG features from the PPG waveform; the PPG features being those features which bear a direct relation with the physiological parameters. In such conventional techniques, the PPG features extracted from the PPG waveform can determine the accuracy of measurement and, therefore, the monitoring of the physiological parameters. Usually, the conventional techniques deploy correlation-based approaches which assess the correlation of the physiological parameters with known values of the physiological parameters, for selecting the PPG features to be extracted. However, the selection of the PPG features based on such approaches may not provide accurate results while monitoring and measuring the physiological parameters. In addition, the set of selected PPG features to be processed and analyzed for monitoring the physiological parameter may still be considerably large and may involve the usage of large amount of computational and temporal resources. [0010] The present subject matter describes methods and devices to monitor physiological parameters associated with a subject using a physiological parameter monitoring device. According to an aspect, the present subject matter involves implementation of photoplethysmographic (PPG) techniques combined with selection of certain relevant PPG features for use in the device for monitoring physiological parameters. According to an aspect of the present subject matter, the selection of the PPG features can involve a two-step approach. In the first step, a correlation between the PPG features and actual known values of the physiological parameter, referred to as ground truth values, is determined. In the second step, the relevant PPG features can be selected based on the strength of correlation between the PPG feature and the ground truth values of the physiological parameter. The physiological parameters can include, for example, heart or pulse rate, pulse oximetry (SpO2) which indicates blood oxygen level, respiration rate, blood pressure, or heart condition based on an electrocardiograph (ECG) features. In an example, the device can be a hand held device, such as a smart phone or a tablet personal computer (PC). [0011] According to an implementation, as part of selection of the relevant PPG features, a video of a body part of a sample subject is captured using a sampling device, for example, a camera. In another case, the sampling device can be a camera of a hand held device, such as a smart phone or a tablet personal computer (PC). In said example, the video of the body part is captured while the body part is abutted against a lens of the camera. For instance, a video of a finger tip or an ear lobe of the subject can be captured. Further, the video is processed to obtain a sample PPG waveform. In one example, the sample PPG waveform can be obtained by processing the video for quantized colour value of each frame and then determining a frequency of the quantized colour value in each frame for a predetermined set of frames. In said example, the sample PPG waveform is then determined based on the frequency of the frames in the predetermined set [0012] Once the PPG waveform is determined, a plurality of sample PPG features is extracted from the sample PPG waveform. In an example, the sample PPG features can be extracted in time domain; however, in another example, the sample PPG features can be extracted in frequency domain. In yet another example, the sample PPG features can be extracted in the time domain as well as frequency domain. In an example, the sample PPG features extracted in the time domain, also referred to as time domain features, can include a peak-to-peak time interval for the peaking frequencies in the sample PPG waveform, pulse interval, crest time indicative of the time taken for the sample PPG waveform to reach the peaking frequencies, delta time, trough to notch time, falling time, notch to trough time, rising slope, falling slope, and area under the cycle. Further, in an example, the sample PPG features extracted in the frequency domain, also referred to as frequency domain features, can include location of peak frequency, distance between the dominant peak frequency and the immediate peak frequency, spectral centroid, and width of dominant peak frequency region. [0013] According to an aspect of the present subject matter, once the PPG features have been extracted, the two-step approach is followed for selecting the relevant PPG features from the entire set of extracted PPG features. As part of selection of the relevant PPG features, the entire set of extracted features can be divided in to one or more training sets and a testing set. In an example, the relevant PPG features can be extracted from the training set, whereas the testing set can be used for determining the relevance of the selected PPG features and the accuracy of the selection. in the training phase, the PPG features and the ground truth values of the physiological parameters are known, and on the basis of the PPG features and the ground truth values, values of the correlation coefficient for each PPG feature is determined. The value of the correlation coefficient of a PPG feature is then used to determine the gain factor for that PPG feature. In an example, a gain function curve can be used for determining the value of the gain factor. In said example, a slope of the gain function curve can be tuned for determining an optimal value of the gain factor for each PPG feature. The optimal gain factors so obtained are used in the testing phase. In the training phase, the PPG features are multiplied by their optimal gain factors and then used for training classifier models for estimating the physiological parameter. On the other hand, during testing, the optimal gain factors can be multiplied by the respective PPG features to estimate the physiological parameters. [0014] Accordingly, in an implementation, a correlation coefficient for each of the plurality of PPG features in the training set, based on the PPG features and the ground truth values. The correlation coefficient can capture a relation between the PPG feature and the ground truth value of the physiological parameter. In an example, the correlation coefficient can be a maximum information coefficient (MIC) value and can be determined based on the MIC techniques. Once the MIC values of the PPG features are determined, strength of the correlation of between each PPG feature and the ground truth values can be determined. Accordingly, a gain factor for each of the plurality of PPG features can be determined, based on the correlation coefficient and a gain function. In an example, the gain function can be a sigmoid gain function. [0015] As would be understood, the gain function, and therefore, the gain factor, can emphasize or highlight the PPG features for which the strength of correlation is high, say based on a threshold value of the MIC value of the PPG feature. Accordingly, in an implementation, each PPG feature is multiplied with the respective gain factor for selecting the relevant PPG features. In an example, the PPG features can be selected based on a threshold value of the gain factor. In another case, the PPG features can be selected based on a threshold value of the PPG feature. In both the above cases, when the PPG feature is multiplied to the gain factor having a low value, say below the threshold value of the gain factor, the value of the PPG feature is suppressed, i.e., falls below the threshold value of the PPG feature, and such PPG features can be discarded. Accordingly, the PPG features for which the value is greater than the threshold value, or for which the value of the gain factor is greater than the threshold value, can be selected as the relevant PPG features. [0016] Subsequently, the testing of the selected relevant features is carried out using the testing set, say previously selected from among the extracted PPG features. In an implementation, the gain factor selected for each PPG feature is employed with the PPG features in the testing set for testing whether the PPG features selected as relevant based on the gain factor are accurately selected or not. In an example, the PPG features in the testing set can be multiplied with the respective gain factors determined for the training set. Based on the multiplication, it can be determined whether the same PPG features are selected as the relevant PPG features from the testing set, as those selected from the training set. [0017] Further, according to an implementation, the relevant PPG features selected above are deployed for estimating and monitoring the physiological parameter in real time. In an example, the relevant PPG features can be used for estimating the ground truth values for the physiological parameter from a PPG waveform and PPG features. In said implementation, the relevant PPG features can be provided on the physiological parameter monitoring device for monitoring the physiological parameter associated with a test subject. [0018] In an implementation, for monitoring the physiological parameter using the physiological parameter monitoring device having the relevant PPG features deployed therein, a video of the test subject can be captured using a camera of the device. In an implementation, the video can be subsequently processed by the physiological parameter monitoring device to obtain a test PPG waveform from which the test PPG features are extracted. In one example, the test PPG waveform is obtained from the video in the same manner as described above for obtaining the sample PPG waveform. In addition, the test PPG features can be the same as the sample PPG features. In another case, the PPG features corresponding to the relevant PPG features can be extracted. Further, based on the test PPG features and the relevant PPG features, the physiological parameter monitoring device can estimate the physiological parameter and monitor the same. [0019] In an example, the physiological parameter monitoring device can indicate a physiological parameter bin or a range estimated for the physiological parameter. Therefore, in said example, the estimation done based on the selected relevant PPG features can be indicative in nature, instead of being quantitative measurement. In such a case, the estimation in accordance with the present subject matter provides for a methodology by way of which the physiological parameters and conditions of the subject can be monitored, for example, to keep a track of the medical condition of the subject so that appropriate medical aid can be provided to the subject in due time. [0020] With the selection of few relevant sample PPG features from the entire set of PPG features extracted from the video, a reduced yet discriminative set of PPG features is obtained for estimating the physiological parameters. Therefore, the accuracy of estimation of the physiological parameter and the monitoring thereof is considerably high. In addition, since during the estimation of the physiological parameter a less number of features are to be analyzed and processed, the computational resources and time involved in monitoring the physiological parameter are substantially less. Therefore, such a set of selected relevant PPG features can be deployed for monitoring the physiological parameters on devices having low processing capabilities. Consequently, the monitoring of the physiological parameters in accordance with the present subject matter is easily scalable and can be made highly available. Therefore, the present subject matter provides for an accurate monitoring of the physiological parameters and, at the same time, the equipment used for such monitoring can be provided as being portable and easy to handle, say in a hand held device such as a mobile phone. [0021] These and other advantages of the present subject matter would be described in greater detail in conjunction with the following figures. While aspects of described systems and methods for monitoring physiological parameters can be implemented in any number of different computing systems, environments, and/or configurations, the embodiments are described in the context of the following device(s). [0022] Fig. 1 illustrates a feature identification system 100 for facilitating monitoring of physiological parameters associated with a subject, in accordance with an embodiment of the present subject matter. In an implementation, the feature identification system 100, based on photoplethysmographic (PPG) techniques and known values of the physiological parameters, can select a small set of relevant PPG features which can be further used for monitoring physiological parameters in real time. In an example, the feature identification system 100 can be implemented as a workstation, a personal computer, say a desktop computer or a laptop, a multiprocessor system, a network computer, a minicomputer, or a server. [0023] In one implementation, the feature identification system 100 includes processor(s) 102 and memory 104. The processor 102 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, and/or any devices that manipulate signals, based on operational instructions. Among other capabilities, the processor(s) is provided to fetch and execute computer-readable instructions stored in the memory 104. The memory 104 may be coupled to the processor 102 and can include any computer-readable medium known in the art including, for example, volatile memory, such as Static Random Access Memory (SRAM) and Dynamic Random Access Memory (DRAM), and/or non-volatile memory, such as Read Only Memory (ROM), erasable programmable ROM, flash memories, hard disks, optical disks, and magnetic tapes. [0024] Further, the feature identification system 100 may include module(s) 106 and data 108. The modules 106 and the data 108 may be coupled to the processors 102. The modules 106, amongst other things, include routines, programs, objects, components, data structures, etc., which perform particular tasks or implement particular abstract data types. In addition, the modules 106 may be implemented as signal processor(s), state machine(s), logic circuitries, and/or any other device or component that manipulate signals based on operational instructions. [0025] In an implementation, the module(s) 106 include a processing module 110, a feature selection module 112, a testing module 114, and other module(s) 116. The other module(s) 116 may include programs or coded instructions that supplement applications or functions performed by the feature identification system 100. Additionally, in said implementation, the data 108 includes a processing data 118, a feature data 120, and other data 122. The other data 122 amongst other things, may serve as a repository for storing data that is processed, received, or generated, as a result of the execution of one or more modules in the module(s). Further, although the data 108 is shown internal to the feature identification system 100, it may be understood that the data 108 can reside in an external repository (not shown in the figure), which may be operably coupled to the feature identification system 100. Accordingly, the feature identification system 100 may be provided with interface(s) (not shown) to communicate with the external repository to obtain information from the data 108. [0026] In addition, for operation, the feature identification system 100 can be coupled to a sampling device 124 to obtain the PPG waveform associated with a sample subject. The feature identification system 100 can select the relevant features from the obtained PPG waveform, for further use in monitoring physiological parameters. Further, the feature identification system 100 interfaces with a physiological parameter monitoring device 130 which uses the selected relevant PPG features to monitor the physiological parameters for a test subject, such as a patient. In an example, the physiological parameter monitoring device 130 can be a hand held device having a processor for providing processing capabilities. For instance, the physiological parameter monitoring device 130 can be a mobile phone, personal digital assistant (PDA), smart phone, or a tablet personal computer. [0027] In operation, the sampling device 124 captures a video of the sample subject for whom ground truth values of a physiological parameter for which the correlation is to be modeled are known. For instance, a body part 126 of the sample subject is positioned on a video capturing device 128, such as a camera, of the sampling device 124. In an example, the sampling device 124 can be a portable video capturing device, such as a smart-phone with video recording capability or a portable video recording camera, i.e., a camcorder. In an example, the subject can position the body part in contact with a lens of the video capturing device 128, or vice-versa, while a flash light of the video capturing device 128 is switched on. For instance, the subject can position a finger tip 128 of his hand on the video capturing device 130 for capturing the video. In another example, the video can be captured from an ear lobe of the subject. [0028] Further, the ground truth values are the actual known values of the physiological parameter. In an example in which the physiological parameter is blood pressure, the ground truth values can be values of systolic blood pressure and diastolic blood pressure. In another example in which the physiological parameter is the ECG features for monitoring heart condition, the ground truth values can be values of ECG features, say QRS complex, PR interval, RR interval, and QT interval. [0029] Subsequently, the sampling device 124 can process the video by dividing the video into frames and obtaining the sample PPG waveform, say based on quantized colour value for each frame. In an implementation, as part of processing of the video, the sampling device 124 can determine one or more quantized colour value for each frame in the plurality of windows. In an example, the sampling device 124 can determine the quantized colour value in the Red-Green-Blue (RGB) colour model and can determine an average value, i.e., the quantized colour value, of any one of the red, blue, or green component for each frame. In another case, the sampling device 124 can determine the average value of any one of the hue, saturation, or value components in the Hue-Saturation-Value (HSV) model, the average value being the quantized colour value. Further, in an implementation, the sampling device 124 can select a set of frames having a predetermined number of frames, for obtaining the sample PPG waveform. In an example, the sampling device 124 can apply Short-Term Fourier Transform (STFT) technique to the quantized colour value of the frames to determine the frequency of the quantized colour value in each frame. Based on the frequency in each frame, the sampling device 124 can obtain an amplitude-time curve for the video; in such a case, the amplitude-time curve being the sample PPG waveform for the set of frames. In another case, the sampling device 124 can obtain an amplitude-frequency curve for the video, in addition to the amplitude-time curve. [0030] In addition, the sampling device 124 can check the captured video for effectiveness, for example, whether the video has sufficient clarity and illumination for being used further. Further, as part of the check, the sampling device 124 can check the video for consistency before obtaining the sample PPG waveform and providing the same to the feature identification system 100. In one implementation, the sampling device 124 can determine the consistency based on a position of peak frequencies of the quantized colour value in a plurality of sets of frames, each set referred to as a window, selected from among all the frames of the video. For example, the sampling device 124 can determine a drift of the peak frequencies across various windows to determine consistency of the video. In said example, if the drift of peak frequency is within a predetermined range, then the video is determined to be consistent. Subsequently, the sampling device 124 selects the set of frames having the predetermined number of frames from the video and determines the PPG waveform from that set, based on the frequency of the quantized colour value in each frame. The sampling device 124 can provide the PPG waveform to the feature identification system 100. Additionally, the sampling device 124 can remove undesired frequencies from the PPG waveform before providing the PPG waveform to the feature identification system 100. For example, the sampling device 124 can determine that a spectrum of PPG waveform can be around 1 Hz. Accordingly, the sampling device 124 can use a 4th order band-pass filter, i.e., cut-off frequencies of around 0.25 Hz and 7 Hz,, and a moving average filter to remove undesired frequencies from the raw PPG waveform. [0031] In another implementation, the sampling device 124 provides the captured video of the sample subject and the processing module 110 of the feature identification system 100 processes the video in the same manner as described above with reference to the sampling device 124, to obtain the PPG waveform. In said implementation, in an example, the sampling device 124 can simply be a video recording device, such as a camera. [0032] Further, the processing module 110 can analyze the sample PPG waveform and obtain a plurality of PPG features from the sample PPG waveform. In an example, the sample PPG features extracted from the sample PPG waveform can include a set of time domain features or a set of frequency domain features, or both. For instance, the set of time domain features can include a peak-to-peak time interval for the peaking frequencies in the sample PPG waveform, pulse interval, crest time indicative of the time taken for the sample PPG waveform to reach the peaking frequencies, delta time, trough to notch time, falling time, notch to trough time, rising slope, falling slope, and area under the cycle. [0033] The determination of the sample PPG features from the PPG waveform by the processing module 110 can be understood with the help of the following illustrations. Consider a case in which the sample PPG features are obtained for determining a model for estimating blood pressure of a subject. In such a case, for obtaining the PPG features, from the sample PPG waveform a systolic peak (Tsn, Asn), a valley point (Tvn, Avn), and a dicrotic notch (Tdn, Adn) are determined, say in the time domain. In said example, T denotes time instant and A denotes the amplitude for the above mentioned features of the sample PPG waveform. For instance, the processing module 110 can determine the systolic peak and the valley point based on local maxima and minima points from the PPG waveform, say a function representative of the PPG waveform. Further, in said example, the processing module 110 can determine the dicrotic notch by first determining a derivative of the function representing the PPG waveform and then identifying a first local maxima between the systolic peak of one PPG waveform and the valley point of the adjacent PPG waveform peak. [0034] Accordingly in said example, the feature of peak to peak interval can be determined as time interval between systolic peaks of two adjacent PPG waveform peaks, the pulse height can be determined as an amplitude of the systolic peak measured from the valley of the PPG waveform, the pulse interval can be measured as time between the valley points of adjacent PPG waveform peaks. In said example, the peak interval, the pulse height, and the pulse interval, and the total area are determined based on the following respective equations: [0035] Peak-to-Peak interval= Tsn+1 – Tsn [0036] Pulse height = Asn – Avn [0037] Pulse interval = Tvn+1 – Tvn [0038] Area under the cycle = , where Pn denotes nth sample value of the PPG waveform. [0039] Furthermore, in an example, the sample PPG feature of crest time can be determined as the time difference between the systolic peak and the valley point of the same PPG waveform peak, the delta time can be measured as a time difference between the dicrotic notch and the systolic peak of the same PPG waveform peak. In said example, the above features can be determined using the following equations as an example: [0040] Crest time = Tsn – Tvn [0041] Delta time = Tdn- Tsn [0042] In addition, the features of trough to notch time, falling time, notch to trough time, rising slope, falling slope can be provided by the following expressions, as examples: [0043] Trough to notch time = Tdn- Tvn [0044] Falling time = Tvn+1- Tsn [0045] Notch to trough time = Tvn+1- Tdn [0046] Rising slope = (Asn – Avn)/(Tsn- Tvn) [0047] Falling slope = (Avn+1 – Asn)/(Tvn+1- Tsn) [0048] Further, the processing module 110 can extract the PPG features in the frequency domain, say from the amplitude-frequency curve. In an example, processing module 110 can extract location of dominant peak frequency, distance between the dominant peak frequency and the immediate peak frequency, spectral centroid, and width of dominant peak frequency region, as the frequency domain features. In an example, for obtaining the frequency domain features, the processing module 110 can segment the frames in the sample video into non-overlapping rectangular windows of 1024 or 256 samples, to obtain sample PPG waveform in the manner as described above. [0049] The processing module 110 can store the PPG features obtained in the time domain and/or in the frequency domain in the processing data 118, and these features can form the entire set of PPG features obtained or extracted from the sample video. [0050] In addition, in an implementation, before the relevant PPG features are selected from the set of PPG features, the processing module 110 can remove intermediate false peaks or trough points from the PPG features to remove noise from the PPG features. Otherwise, actual peaks or trough points may be completely missed out due to noisy surroundings and may result in the incorrect calculation of PPG features during extraction of the PPG features. In an example, the processing module 110 can create two clusters of the PPG features. Further, based on a histogram analysis, the processing module 110 can intitlize the centroids for the cluster analysis. Subsquently, the processing module 110 can apply a 2-Means clustering followed by cluster density estimation to remove the incorrect PPG features. In another case, the processing module 110 can apply k-means algorithm to obtain the cluster centroids. Further, the processing module 110 can employ Xie-Beni index for removing the incorrect PPG features and obtaining the set of PPG features which can be used for selection of the relevant PPG features. [0051] Further, in an implementation, the feature selection module 112 can select one or more relevant sample PPG features from the plurality of sample PPG features. In accordance with an aspect of the present subject matter, the feature selection module 112 can follow a two-step approach for selecting the relevant PPG features from the entire set of extracted PPG features. In the first step, the feature selection module 112 can determine the correlation between the PPG features and the ground truth values of the physiological parameter. Further, in the second step, the features selection module 112 can select the relevant PPG features based on the strength of correlation between the PPG feature and ground truth values of the physiological parameter. [0052] According to an implementation, as part of selection of the relevant PPG features, the feature selection module 112 can divide the entire set of extracted PPG features into one or more training sets and a testing set and store the same in the feature data 120. In an example, the feature selection module 112 can extract the relevant PPG features from the training set, and use the testing set to determine accuracy of the selection of the relevant PPG features. [0053] Accordingly, the feature selection module 112 can determine a correlation coefficient for each of the plurality of PPG features in the training set, based on the PPG features and the ground truth values. The correlation coefficient can capture a relation between the PPG feature and the ground truth value of the physiological parameter. In an example, the feature selection module 112 can determine a maximum information coefficient (MIC) value as the correlation coefficient, based on the MIC techniques. In an example, the feature selection module 112 can construct grids with various sizes to find the largest mutual information between the data pair, i.e., between the PPG feature and the ground truth value. For each pair of data (x, y), if I is the mutual information for a grid G, then MIC of a set D of pairwise data with sample size n and grid size (xy), the feature selection module 112 can determine the correlation coefficient, i.e., the MIC value based on the following relation as an example: MIC(D) = maxxy