We prove that the vast majority of symmetric states of qubits (or spin 1/2) can be decomposed in a unique way into a superposition of spin 1/2 coherent states. For the case of two qubits, the proposed decomposition reproduces the Schmidt decomposition and therefore, in the case of a higher number of qubits, can be considered as its generalization. We analyze the geometrical aspects of the proposed representation and its invariant properties under the action of local unitary and local invertible transformations. As an application, we identify the most general classes of entanglement and representative states for any number of qubits in a symmetric state.Symmetric states under permutations have drawn lately a lot of attention in the field of quantum information. The essential reason is that the number of parameters needed for the description of a state in a symmetric subspace scales just linearly with the number of parties. This simplification makes symmetric states a good testground for complex quantum information tasks such as the description of multipartite entanglement [1-8] and quantum tomography [9].There are two well known representations for symmetric states, the Dicke basis [10] and Majorana representation [11]. In the present work, we introduce a novel representation whose forms resembles strongly to a generalized Schmidt decomposition and that presents advantages with respect to the previous ones as regards the readability of the entanglement properties of the state. The structure of the proposed representation remains invariant under the action of local unitary operations which leave the state in the symmetric subspace. We associate to the proposed decomposition a geometric representation of the states that has the interest of displaying this invariance. We proceed by identifying invariant forms of the representation under the action of local unitary and local invertible transformations. In this way we arrive to an exact classification of entanglement for symmetric states. With the well studied example of three qubits we establish a first connection among the suggested representation and measures of multipartite entanglement. Furthermore, an immediate consequence of our methods is a straightforward estimation of a well established measure of entanglement, the so called Schmidt measure [12] for the symmetric states.The structure of this paper is the following. We start by recalling some basic properties of the Dicke and Majorana representation and their connection to entanglement classification. We then move to the presentation of the novel decomposition that is the main result of the paper. Finally we employ the properties of the decomposition in order to arrive at a classification of entanglement applicable to the vast majority of symmetric states.Every symmetric state of N qubits can be expressed in a unique way over the Dicke basis formed by the N + 1 joined eigenstates {|N/2, m } of the collective operatorŝ This projection leads to a polynomial of N th order on the complex parameter α, the so-calle...