Neural oscillator network modelling of auditory stream segregation

想象你在一个晚宴上。餐桌上正在进行五次谈话，背景音乐嗡嗡作响，餐具叮当作响，玻璃杯叮当作响。陷入一场关于股票期权的无聊讨论中，你希望自己能和身后的人一起回忆在阿尔卑斯山滑雪的情景。无论你在哪里，你的大脑总是在一个声音的世界里嗡嗡作响。我们怎么能选择收听什么节目呢？尽管单个声源（如人声，嗡嗡作响的冰箱）会随着时间的推移而改变，但有些特征是不变的，比如它们来自哪里，音高是高还是低，以及它们重复的频率。在如上所述的情况下，我们可以控制自己听到的声音，我们可以专注于一个人的声音或对话，而把其他声音放在背景中。有证据表明，大脑使用两种策略来区分声源：1)通过特征（例如，声源是高音还是低音，声源来自不同的位置），2)通过时间和节奏（人们以不同的速度说话，在不同的时间开始/停止说话）。大脑的声音处理通路根据声音的特征来区分声音，例如，不同的神经元组对高音和低音做出反应。持续的大脑节律（大脑活动的振荡）可以与特定的声源同步，以便追踪它们。结合在一起，大脑可以通过神经元群跟踪特征，并随着时间的推移同步跟踪这些特征的活动来跟踪特定的声源。动力系统是描述随时间变化的过程（如大脑节律）的数学领域。在其他类型的动力学中，它帮助我们理解了生物、物理、化学、社会网络和技术应用之间的振荡和同步。神经元活动的振荡对很多认知功能都很重要，比如做决定、形成记忆和欣赏音乐。这里的研究重点是利用关于振荡的数学理论来开发一个涉及处理和分离声源的大脑区域的计算机模型。目前计算机建模的技术水平主要集中在第一种策略（按特征分离）上。在这里，我们的重点将是把这个和第二个策略结合起来，特别关注大脑是如何利用时间和节奏来区分声音的。这是基于一个假设，即振荡的同步性对于跟踪声音的时间至关重要。事实上，当我们听到重复的声音（比如简单的音乐节奏）时，大脑听觉和运动区域的神经元都会随着节拍及时启动，即使我们自己没有活动。这项研究旨在揭示这两个区域是如何协同工作的，使我们能够同时跟踪我们关注的声源（股票期权）和我们真正想要关注的声源（在阿尔卑斯山滑雪）。

Imagine yourself at a dinner party. Five conversations are going on around the table, music hums in the background, cutlery rattles and glasses clink. Caught in a dull discussion about stock options, you're wishing you were with the people over your shoulder reminiscing about skiing in the Alps. No matter where you are, your brain is constantly buzzing in a world of sound. How are we able to choose what to tune in to?Although individual sound sources (e.g. a voice, a humming fridge) change over time some features remain constant, like where they're coming from, whether they are high or low in pitch and how often they repeat. In a situation like the one described above we have some control over what we hear, we can focus on an individual voice or conversation whilst pushing other sounds into the background. There is evidence that the brain uses two strategies to differentiate sound sources: 1) by features (e.g. sources are high or low pitch, sources come from different locations), and2) by timing and rhythm (people talk at different speeds and start/stop stop talking at different times).The brain's sound processing pathways separate sounds by their features, resulting in, for example, different groups of neurons responding to high and low pitch sounds. Ongoing brain rhythms (oscillations in brain activity) can synchronise with specific sound sources in order to track them. Combined together this allows the brain to follow specific sound sources with groups of neurons tracking features and synchronisation of their activity tracking these features over time.Dynamical systems is a field of mathematics describing processes (such as brain rhythms) that change over time. Amongst other types of dynamics, it has helped us understand oscillations and synchronisation across biology, physics, and chemistry to social networks and technological applications. Oscillations in the activity of neurons is important for lots of cognitive functions like making decisions, forming memories and enjoying music. The research here focuses on using mathematical theory about oscillations to develop a computer model of the brain regions involved in processing and segregating sound sources. The current state of the art in computer modelling has focused on the first strategy (separating by features). Here, the focus will be on integrating this with the second strategy, paying specific attention to how the brain uses timing and rhythm to segregate sounds. This is based on the hypothesis that synchronisation of oscillations is crucial for tracking sounds over time. Indeed, when we listen to repetitive sounds (like a simple musical rhythm) neurons in both auditory and motor regions of the brain start to fire in time to the beat, even when we aren't moving ourselves. The research aims to reveal how both of these regions working together enable us simultaneously keep track of both the sound source we're focusing on (stock options) and the one we'd like to really be paying attention to (skiing in the Alps).