Sahu Nagesh Kumar, Mehar Varsha; International Journal of Advance Research, Ideas and Innovations in Technology © 2018, www.IJARIIT.com All Rights Reserved Page | 134 ISSN: 2454-132X Impact factor: 4.295 (Volume 4, Issue 5) Available online at: www.ijariit.com A review of signal parameter estimation techniques Nagesh Kumar Sahu raku.mandal80@gmail.com Bhabha College of Education, Bhopal, Madhya Pradesh Varsha Mehar atulmandal93@gmail.com Bhabha College of Education, Bhopal, Madhya Pradesh ABSTRACT Signal Investigation, the signals to be detected usually include unidentified parameters such as amplitude, time delay, phase, and frequency; these parameters must be probable previous to the signal discovery. The techniques used to approximation these signal parameters can be broadly confidential into two main categories known as parametric and non-parametric methods. This paper presents a review of these signal parameter estimation techniques. Keywords— Parametric and Non-Parametric methods, Signal parameter estimation 1. INTRODUCTION Signal stricture opinion, and hence uncovering, problems are worried about the examination of received signals to conclude the deficiency or incidence of a signal of concern; the removal of in the order in these signals as well as the signal categorization [1]. These are problems of meaning in applications such as seismic examination, speech appreciation, cellular mobile message, biomedical manufacturing, radar, and sonar signal dispensation. The signals to be detected often contain unidentified parameters, such as amplitude, time delay, phase, and frequency; these parameters must be estimated before any signal detection. To estimate these parameters, a figure of methods can be useful. Generally, signal parameter judgment techniques can be classified into two main categories; namely parametric and non-parametric approaches [2-4]. An appraisal of several of these signal parameter estimation methods is presented in this paper. 2. PARAMETRIC AND NON-PARAMETRIC ESTIMATION TECHNIQUE In this paper, Non-parametric techniques are Fourier-based methods of providing ghostly estimates anywhere no previous replica is unspecified, in the intellect that no assumption is complete about the bodily procedure that generates a given data. They are also recognized as the traditional methods of ghostly opinion. Although this advance of signal stricture opinion is computationally competent, it though has unfinished promptness declaration. These methods also suffer from ghostly leakage belongings that frequently mask weak signals. Prominent conclusions from these non-parametric techniques are that there is forever a concession in the bias-variance trade- off because both of these errors cannot be minimized concurrently [2-5]. The parametric-based method can although be old to take out a high-resolution estimate, particularly in an application where small data minutes are obtainable due to the temporary phenomenon, provide the signal arrangement is known. These techniques are also recognized as a model-based method of ghostly opinion, where a generated model with documented useful form is unspecified. The parameter in the unspecified model is then predictable, and a signal’s ghostly individuality of notice resulting from the predictable model. Therefore, the predictable spectral individuality is only as good as the fundamental model. example of these parametric-based technique comprises the autoregressive (AR) process model (comprise the Yule-Walker [4, 6] and least squares method [4] ), the moving average (MA) course model, as well as the joint autoregressive moving average (ARMA) process model [2-4]. The autoregressive process model, such as the Prony algorithm, is the simplest of the parametric-based techniques. The Prony algorithm, which models sample data as a linear mixture of exponentials, is a method that can be used to identifying the frequencies, amplitudes, and phases of a signal. Even though the Prony algorithm has the ability to resolve rays much closer than the Fourier-based limit, it, however, has a tendency to yield biased estimates. A further instance, of the parametric-based method, is the space-alternating general expectation-maximization (SAGE) algorithm [7]. The SAGE algorithm is a low-complexity generalization of the expectation-maximization (EM) algorithm [8]. The EM algorithm is an iterative process used to calculate an utmost likelihood approximation when an observed data is regarded as unfinished [9]. The SAGE algorithm breaks down a multi-dimensional optimization procedure, essential to calculate the estimates of the parameter of a gesture, into more than a few separate, low-dimensional maximization events, which are performed in sequence; thus reducing the computational cost. In addition, this algorithm overpowers the decree restraint intrinsic in the Fourier-based method. Though, the SAGE algorithm depends on the supposition that a limited recognized number of waves characterize by their spread delay, complex amplitude, and azimuthally incidence way is impinge in the neighborhood of a receiver. Underestimate the number of impinge waves can result in poor resolution, while over-estimation can give rise to spurious components in the parameter estimates.