We often think of vectors as arrows with a specific length and direction, existing in a 2D or 3D space. But in functional analysis, the concept of a vector is much broader. A function, like a continuous audio signal over time, can also be treated as a vector. Instead of having a few components (like x, y, and z), its components are its values at every single point in time. This means it lives in an infinite-dimensional vector space.

Not just any function will do. For signal processing, we work with functions in a specific kind of vector space called a Hilbert space named $L^2$. This space contains all complex-valued, square-integrable functions. The 'square-integrable' part is crucial. It means that the total energy of the signal is finite. A signal with infinite energy would be physically impossible to create or measure.

In familiar 2D space, the dot product tells us how much one vector points in the direction of another. A large positive dot product means they are aligned; zero means they are orthogonal. We can extend this idea to the infinite-dimensional world of signals using the inner product.

The inner product of two signals, $f(t)$ and $g(t)$, quantifies their 'overlap' or 'similarity'. If two signals are very similar, their inner product will be large. If they are completely dissimilar (orthogonal), their inner product will be zero. This operation is the key to breaking down a complex signal into its constituent frequencies.

The inner product has a special case: taking the inner product of a signal with itself. This gives us the squared 'norm' or 'length' of the signal vector. In signal processing, this value has a direct physical meaning: it is the total energy of the signal.

The norm itself, $||f||$, is the square root of the energy and is analogous to the length of a geometric vector. It's often called the Root Mean Square (RMS) value of the signal. By defining signals as vectors in a Hilbert space, we gain access to all the geometric tools of linear algebra, like projections, orthogonality, and norms. This framework is what allows us to precisely extract frequency information from complex signals, which is the basis of Fourier analysis.