The Meaning of Convolution Integrals in Memory Systems | 卷积积分在记忆性系统中的意义

📚 The Meaning of Convolution Integrals in Memory Systems | 卷积积分在记忆性系统中的意义

In many natural and engineered systems, the present output is not decided only by the present input; it also carries a memory of the past. The convolution integral is the mathematical language that describes how such a system accumulates past influences and translates them into a current response. In IB mathematics, where calculus and modelling meet, the convolution integral connects differential equations, signal processing, and real-world problems.

在许多自然与工程系统中，当前的输出并不只由当前的输入决定，它还可能携带对过去的记忆。卷积积分正是用来描述此类系统如何累积过去影响、并将其转化为当前响应的数学语言。在 IB 数学中，当微积分与建模相遇时，卷积积分将微分方程、信号处理以及现实问题联系起来。

---

1. What Is a Convolution Integral? | 什么是卷积积分？

For two functions f and g, the convolution is written as f∗g and defined by the integral

(f∗g)(t) = ∫_−∞^∞ f(τ)g(t−τ)dτ

Here τ is a dummy variable of integration. The product f(τ)g(t−τ) is computed for every time shift τ, and then the results are added over all possible shifts. The output function g appears reversed and moving through time, which is why the operation is sometimes called “folding and shifting.”

对于两个函数 f 和 g，卷积写作 f∗g，并由以下积分定义：

(f∗g)(t) = ∫_−∞^∞ f(τ)g(t−τ)dτ

这里的 τ 是积分变量。我们对每一个时间位移 τ 计算乘积 f(τ)g(t−τ)，然后对所有可能的位移求和。函数 g 被视为“翻转后平移”地扫过时间轴，因此卷积有时也被称为“翻转与平移”运算。

---

2. Memory Systems and Their Input-Output Law | 记忆性系统及其输入输出规律

A system is said to have memory when its output at time t depends on input values from the past, not only on the present value. For example, a capacitor remembers the history of the current flowing through it; a moving object remembers the forces that acted on it earlier because of momentum; and an economic model may remember previous interest rates when forecasting future inflation.

如果一个系统在时刻 t 的输出依赖于过去时刻的输入值，而不仅仅取决于当前输入值，那么该系统就具有记忆性。例如，电容器会记住流过它的电流的历史；运动物体因动量而记住此前作用在它身上的力；经济模型在预测未来通胀时，也可能记住先前的利率水平。

For a linear system, doubling the input doubles the output, and the response to a sum of inputs is the sum of the responses. For a time-invariant system, delaying the input by some amount simply delays the output by the same amount. A linear, time-invariant system with memory is completely characterized by one special function: its impulse response. The convolution integral then provides the rule that turns this response into the output for any input.

对于线性系统，输入加倍会使输出加倍，并且对输入之和的响应等于各响应之和。对于时不变系统，输入延迟一段时间，输出也会延迟相同的时间。一个具有记忆性的线性时不变系统，完全由某个特殊函数刻画：即它的冲激响应。卷积积分正是利用这个函数，对任意输入求出对应输出的规则。

---

3. The Impulse Response: The System’s Memory Trace | 冲激响应：系统的记忆痕迹

The impulse response h(t) is the output of the system when the input is a single idealized spike, called the Dirac delta function δ(t). This spike lasts for an instant and has total area one. If the spike is sent into the system at time τ, then because the system is time-invariant, the response is the same shape, shifted to time τ: h(t−τ).

冲激响应 h(t) 是指当输入为单个理想化尖峰，即狄拉克δ函数 δ(t) 时，系统的输出。这个尖峰只持续一瞬间，并且总面积为 1。如果该尖峰在时刻 τ 进入系统，由于系统具有时不变性，响应形状不变，只是平移到时刻 τ，即 h(t−τ)。

The impulse response tells us how strongly the system remembers an input that happened τ units ago. If h(t−τ) is large, the past input still has a strong effect at time t. If h(t−τ) is nearly zero, the memory of that input has almost disappeared. In this sense, h is the “memory trace” of the system.

冲激响应告诉我们，系统对发生在 τ 时间之前的输入究竟记得多牢。如果 h(t−τ) 很大，说明该过去输入在时刻 t 仍有较强影响；如果 h(t−τ) 几乎为零，说明系统对这个输入的记忆已经基本消失。因此，h 可以被理解为系统的“记忆痕迹”。

---

4. Deriving Convolution from Superposition | 从叠加原理推导卷积

Any input function x(t) can be imagined as a continuous sequence of very narrow impulses. Mathematically, the sifting property of the delta function writes x(t) as

x(t) = ∫_−∞^∞ x(τ)δ(t−τ)dτ

Each term x(τ)δ(t−τ) represents a small impulse with strength x(τ) located at time τ. Because the system is linear, the response to this small impulse is x(τ)h(t−τ). Summing the responses over all τ gives the output:

任何输入函数 x(t) 都可以被想象成一系列非常窄的冲激序列。根据 δ 函数的筛选性质，x(t) 可以写成：

x(t) = ∫_−∞^∞ x(τ)δ(t−τ)dτ

其中每一项 x(τ)δ(t−τ) 都表示位于时刻 τ、强度为 x(τ) 的微小冲激。由于系统是线性的，系统对这个微小冲激的响应为 x(τ)h(t−τ)。把所有 τ 对应的响应叠加起来，就得到输出：

y(t) = ∫_−∞^∞ x(τ)h(t−τ)dτ

This is exactly a convolution. It shows that the output at time t is the sum of all past inputs, each weighted by the system’s memory h(t−τ). This is the central reason why convolution integrals appear when studying systems with memory.

这正是卷积。它表明时刻 t 的输出，是所有过去输入按系统的记忆 h(t−τ) 加权后得到的总和。这就是为什么在研究具有记忆性的系统时，卷积积分会成为核心工具。

---

5. Causality and the Limits of Integration | 因果性与积分限

A system is causal if its output cannot depend on future inputs. For a causal system, the impulse response h(u) must be zero for u < 0, because the output cannot respond before the impulse has arrived. Applying this condition to the convolution integral changes the upper limit from ∞ to t:

如果一个系统的输出不可能依赖于未来输入，那么该系统就是因果系统。对于因果系统，冲激响应 h(u) 在 u < 0 时必须为零，因为在冲激到达之前不可能产生响应。将这个条件代入卷积积分，就可以把上限从 ∞ 改为 t：

y(t) = ∫_−∞^t x(τ)h(t−τ)dτ

If the input also begins at time 0, then x(τ)=0 for τ<0, and the lower limit becomes 0. The result is the familiar "memory integral"

如果输入从时刻 0 才开始，那么在 τ<0 时有 x(τ)=0，于是下限也变为 0。于是我们得到常见的“记忆积分”形式：

y(t) = ∫₀^t x(τ)h(t−τ)dτ

The interval [0,t] is exactly the portion of history that can influence the present. This is why the convolution integral is not only a mathematical operation but also a representation of causality in physical and engineered systems.

区间 [0,t] 正是能够影响当前状态的“历史窗口”。因此，卷积积分不仅是一种数学运算，更是物理系统和工程系统中因果性的体现。

---

6. Fading Memory and Weighting | 衰减记忆与加权

In most memory systems, older inputs have a weaker influence than recent inputs. This is called fading memory, and it is encoded in the decay of h(t). A simple and important example is the exponential memory kernel

在大多数记忆性系统中，较早输入的影响比近期输入更弱，这被称为“衰减记忆”，它通过 h(t) 的衰减来体现。一个简单而重要的例子是指数记忆核：

h(t) = (1/T)e^−t/T, t ≥ 0; h(t) = 0, t < 0Published by TutorHao | IB Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）