📚 Year 11 WJEC Computer Science: Speaking and Listening Exam Prep | Year 11 WJEC 计算机：口语/听力备考专项
This article focuses on the 'speaking and listening' aspects of computer science – a fascinating area that covers how machines process human speech. In the WJEC GCSE Computer Science course, you will encounter fundamental concepts of sound representation, voice interfaces, and the principles behind speech recognition and synthesis. Grasping these ideas is essential not only for exams but for understanding everyday technology like voice assistants. Let's explore the key topics and effective revision strategies to help you succeed in this part of the specification.
本文聚焦于计算机科学中的“口语与听力”领域——一个关于机器如何处理人类语音的迷人方向。在 WJEC GCSE 计算机科学课程中，你将接触到声音的表示、语音界面，以及语音识别与合成背后的基本原理。掌握这些概念不仅对考试至关重要，也能帮助你理解语音助手等日常技术。让我们一同梳理核心主题和高效的复习方法，助你在这一部分取得优异表现。
1. Sound as Data: Sampling and Quantisation | 声音即数据：采样与量化
Computers are digital machines, but sound waves are analogue. To store and process audio, a computer must convert continuous sound into a stream of binary numbers. This process involves two key steps: sampling and quantisation. The sample rate is how many times per second the amplitude of the sound wave is measured, measured in hertz (Hz). The higher the sample rate, the more accurately the original wave is captured, but the larger the file size becomes.
计算机是数字设备，而声波是模拟的。为了存储和处理音频，计算机必须将连续的声音转换为二进制数据流。这个过程包含两个关键步骤：采样和量化。采样率是指每秒对声波振幅进行测量的次数，单位为赫兹（Hz）。采样率越高，原始声波的还原度就越高，但同时文件体积也会越大。
A typical CD-quality audio uses a sample rate of 44 100 Hz, meaning 44 100 samples are taken each second. Each sample is then quantised – assigned a binary value from a limited set of levels. The number of bits used for each sample (bit depth) determines the dynamic range. A higher bit depth reduces quantisation noise and improves fidelity, but again increases storage requirements.
典型的 CD 品质音频使用 44 100 Hz 的采样率，即每秒采集 44 100 个样本。每个样本随后被量化——从有限的级别中分配一个二进制数值。每个样本使用的比特数（位深度）决定了动态范围。更高的位深度可减少量化噪声并提升保真度，但同样会增加存储需求。
2. Audio File Size Calculation | 音频文件大小计算
te (Hz) × bit depth (bits) × number of channels × duration (seconds). Remember to convert bits to bytes by dividing by 8, and further to kilobytes or megabytes as needed.
计算音频文件大小是 WJEC 考试中常见的技能。公式很简单：文件大小（比特）= 采样率（Hz）× 位深度（比特）× 声道数 × 时长（秒）。请记住，要将比特转换为字节需除以 8，并可按需进一步转换为千字节或兆字节。
File Size (bits) = Sample Rate × Bit Depth × Channels × Time
0 = 42 336 000 bits ≈ 5.04 MB. Practice such calculations until they become second nature.
例如，一段 30 秒的立体声音频，以 44.1 kHz 采样率、16 位深度录制，文件大小计算为：44 100 × 16 × 2 × 30 = 42 336 000 比特 ≈ 5.04 MB。多加练习，直到这类计算成为你的第二本能。
3. From Sound to Phonemes: The Speech Recognition Pipeline | 从声音到音素：语音识别流水线
Speech recognition, or 'listening', is the process by which a computer interprets spoken language. It begins with an acoustic signal captured by a microphone. The signal is digitised and then broken into small frames (typically 10-25 ms). Each frame is analysed to extract features such as MFCCs (Mel-frequency cepstral coefficients) that represent the vocal tract shape at that instant.
语音识别，即“听力”，是计算机解析口头语言的过程。它从麦克风捕捉的声音信号开始。信号被数字化后，被切分成小帧（通常为 10-25 毫秒）。每一帧都会被分析，以提取诸如梅尔频率倒谱系数（MFCC）等特征，这些特征代表了该瞬间的声道形状。
These feature vectors are then fed into an acoustic model, often a Hidden Markov Model (HMM) or a deep neural network, which maps sound patterns to basic speech units called phonemes. Phonemes are the smallest distinctive units of sound in a language, such as /p/, /k/, or /i:/. A language model then predicts likely word sequences, using statistical patterns or neural networks to disambiguate similar-sounding phrases.
这些特征向量随后被输入声学模型（通常是隐马尔可夫模型 HMM 或深度神经网络），该模型将声音模式映射到称为音素的基本语音单位。音素是语言中最小的辨义声音单位，如 /p/、/k/ 或 /iː/。然后，语言模型会利用统计模式或神经网络预测可能的词语序列，以消除同音短语的歧义。
4. Speech Synthesis: How Machines Talk | 语音合成：机器如何说话
Speech synthesis, or 'speaking', is the reverse process. It converts text into artificial speech. The most advanced systems use concatenative synthesis or parametric neural methods. Concatenative synthesis stitches together pre-recorded fragments of real speech (diphones or whole words), while modern neural TTS (Text-to-Speech) uses models like WaveNet to generate raw audio waveforms directly from text, producing very natural-sounding voices.
语音合成，即“说话”，是相反的过程。它将文本转换为人工语音。最先进的系统使用拼接合成或参数化神经方法。拼接合成将预先录制好的真实语音片段（双音子或整个词）拼接起来，而现代的神经文本转语音（TTS）则利用 WaveNet 等模型直接从文本生成原始音频波形，产生非常自然的说话声音。
In the WJEC syllabus, you are expected to understand the generic TTS process: text pre-processing (normalising numbers, abbreviations), letter-to-sound conversion using rules or dictionaries, and prosody generation (intonation, stress, rhythm) to make the output sound human-like rather than robotic.
在 WJEC 教学大纲中，你需要了解通用的 TTS 流程：文本预处理（规范化数字、缩写），使用规则或词典进行字母到声音的转换，以及韵律生成（语调、重音、节奏），使得输出语音听起来更像人类而非机器人。
5. Input and Output Devices for Voice | 语音输入与输出设备
While studying 'speaking and listening' in computing, we must not forget the hardware. A microphone is an input transducer that converts sound pressure waves into an electrical analogue signal. The computer's sound card then converts this analogue signal into digital using an Analogue-to-Digital Converter (ADC). On the other side, speakers or headphones act as output transducers. A Digital-to-Analogue Converter (DAC) reconstructs the digital audio stream back into an analogue signal that vibrates the speaker cone to produce sound waves.
在学习计算机的“口语与听力”时，我们不能忘记硬件。麦克风是一种输入换能器，它将声压波转换为电模拟信号。计算机的声卡随后利用模数转换器（ADC）将该模拟信号转换为数字信号。另一方面，扬声器或耳机充当输出换能器。数模转换器（DAC）将数字音频流重新构建为模拟信号，驱动扬声器振膜振动以产生声波。
Key factors affecting quality include the microphone's frequency response, the ADC's sample rate and bit depth, and the DAC's linearity and noise floor. These are all relevant when discussing sound fidelity in exam questions.
影响品质的关键因素包括麦克风的频率响应、ADC 的采样率和位深度，以及 DAC 的线性度和底噪。在考试中讨论声音保真度时，这些内容都息息相关。
6. Data Compression for Voice: Lossy vs Lossless | 语音数据压缩：有损与无损
Voice data can be very large, so compression is vital for streaming and storage. Lossless compression (e.g., FLAC, ALAC) preserves every bit of the original sound, allowing exact reconstruction. However, it achieves only modest compression ratios. Lossy compression (e.g., MP3, AAC, Ogg Vorbis) removes audio data that is considered perceptually irrelevant, based on psychoacoustic models that exploit the limitations of human hearing.
语音数据可能非常庞大，因此压缩对传输和存储至关重要。无损压缩（如 FLAC、ALAC）保留了原始声音的每一个比特，允许精确重构。然而，它只能达到有限的压缩比。有损压缩（如 MP3、AAC、Ogg Vorbis）则基于心理声学模型，利用人耳听觉的局限性，去除那些被认为感知上不重要的音频数据。
For GCSE, you should know that lossy formats achieve much smaller file sizes with little perceived loss in quality. Compression algorithms like MP3 use techniques such as perceptual noise shaping and masking (where a loud sound covers up a quieter sound of similar frequency). This is why voice assistants can transmit speech efficiently over the internet.
对于 GCSE，你应该知道有损格式可以大幅缩小文件体积，且人耳几乎感知不到质量损失。像 MP3 这样的压缩算法使用了感知噪声整形和掩蔽技术（一个响亮的声音会掩盖频率相近的较轻声音）。这就是语音助手能高效地在互联网上传输语音的原因。
7. Speech in Human-Computer Interaction (HCI) | 语音在人机交互中的应用
Voice User Interfaces (VUIs) allow users to interact with systems through voice commands. From smart speakers like Amazon Echo to Siri on an iPhone, principles of HCI apply particularly to designing systems that are intuitive and error-tolerant. A key challenge is handling varied accents, background noise, and homophones. Error recovery strategies, such as confirmation prompts ('Did you mean...?'), are crucial for usability.
语音用户界面（VUI）允许用户通过语音命令与系统交互。从 Amazon Echo 这样的智能音箱到 iPhone 上的 Siri，人机交互的原则在设计直观且容错的系统时尤为重要。一个关键挑战是处理不同的口音、背景噪音和同音词。错误恢复策略，如确认提示（“您的意思是...？”），对于可用性至关重要。
For your exam, be able to discuss how voice recognition technology is a form of biometric input, since voice patterns are unique to individuals. Also understand the ethical considerations: privacy (always-on microphones), data security, and accessibility for those with disabilities.
对于考试，要能够讨论语音识别技术如何构成一种生物特征输入，因为语音模式对每个人来说都是独一无二的。同时，要理解相关的伦理考量：隐私（始终开启的麦克风）、数据安全以及为残障人士提供的无障碍性。
8. Binary Representation of Text: ASCII and Unicode | 文本的二进制表示：ASCII 与 Unicode
Speech systems ultimately process text both before synthesis and after recognition. To represent characters internally, computers use standardised codes. ASCII uses 7 or 8 bits per character, covering English letters, digits, and symbols. However, it cannot represent diverse scripts like Chinese or emoticons. Unicode (UTF-8, UTF-16) uses a variable number of bytes to encode over 140 000 characters from all languages, making it essential for global voice applications.
语音系统在合成之前和识别之后最终都要处理文本。为了在内部表示字符，计算机使用标准化的编码。ASCII 为每个字符使用 7 或 8 个比特，涵盖了英文字母、数字和符号。然而，它无法表示中文或表情符号等多种文字。Unicode（UTF-8、UTF-16）使用可变字节数来编码超过 14 万个来自所有语言的字符，这使其成为全球语音应用的必需品。
You may be asked to compare ASCII and Unicode, or to explain why a speech-to-text system must support Unicode to handle multilingual input. Remember that storage and transmission efficiency matters: UTF-8 is backward compatible with ASCII for English text.
考试可能会要求你比较 ASCII 和 Unicode，或者解释为什么语音转文本系统必须支持 Unicode 以处理多语言输入。请记住，存储和传输效率很重要：对于英文文本，UTF-8 向后兼容 ASCII。
9. Common Exam Question Types and Model Answers | 常见考试题型与范例答案
WJEC exam papers often ask you to describe how sound is stored digitally, calculate file sizes, or explain the purpose of components in a speech recognition system. A typical 4-mark question might be: 'Describe how an analogue sound wave is converted into a digital format.' Your answer should include sampling, quantisation, and the role of ADC. Use diagrams if prompted — boxes with labels can earn marks.
WJEC 试卷常会要求你描述声音如何以数字方式存储、计算文件大小，或解释语音识别系统中各组件的作用。一道典型的 4 分题可能是：“描述模拟声波如何转换为数字格式。”你的答案应包括采样、量化以及 ADC 的作用。如有提示，可使用图示——带有标签的方框图能为你赢得分数。
A model answer: 'The microphone captures sound as an analogue electrical signal. The ADC samples the signal at regular intervals (e.g., 44 100 times per second). Each sample’s amplitude is measured and rounded to a discrete value based on the bit depth. This series of binary numbers represents the digital audio.' Always structure answers logically and use subject terminology.
范例答案：“麦克风将声音捕捉为模拟电信号。模数转换器以固定间隔（例如每秒 44 100 次）对信号进行采样。测量每个样本的振幅，并根据位深度将其舍入为离散值。这一系列二进制数就代表了数字音频。”始终要有逻辑地组织答案，并使用学科术语。
10. Practical Revision Activities for Speaking/Listening Topics | 口语/听力话题的实践复习活动
Active revision is far more effective than passive reading. Try creating a flashcard set with terms like 'sample rate', 'bit depth', 'phoneme', 'MFCC', 'ADC/DAC', 'lossy compression'. On the other side, write the definition and an example. Use online tools to record your own voice and examine the waveform; notice how plosive sounds ('p', 'b') create spikes.
主动复习远比被动阅读有效。尝试制作一套抽认卡，包含“采样率”、“位深度”、“音素”、“MFCC”、“ADC/DAC”、“有损压缩”等术语。在卡片的另一面写上定义和一个例子。使用在线工具录制自己的声音并观察波形；留意爆破音（“p”、“b”）是如何产生尖峰的。
Teaching a friend the process of speech recognition from microphone to text output is also an excellent way to solidify knowledge. Practise drawing annotated flowcharts. Past paper questions, especially those from the WJEC specification, will reveal the depth of detail expected. Time yourself and mark using the official mark schemes.
向朋友讲解从麦克风到文本输出的语音识别过程，也是巩固知识的绝佳方法。练习绘制带有注释的流程图。历年真题，特别是 WJEC 考试局的题目，会揭示期望的详细程度。计时作答，并使用官方评分标准为自己打分。
11. Key Vocabulary Checklist | 关键词汇清单
| English Term | 中文术语 | Brief Explanation |
| Analogue Signal | 模拟信号 | A continuous wave that varies smoothly over time. |
| Digital Signal | 数字信号 | A discrete representation using binary values. |
| Sample Rate | 采样率 | Number of samples taken per second (Hz). |
| Bit Depth | 位深度 | Number of bits used to represent each sample. |
| ADC / DAC | 模数/数模转换器 | Converts analogue to digital, and vice versa. |
| Phoneme | 音素 | Smallest unit of sound in speech. |
| MFCC | 梅尔频率倒谱系数 | Features representing the short-term power spectrum of sound. |
| Lossy Compression | 有损压缩 | Reduces file size by removing non-essential data permanently. |
| TTS | 文本转语音 | Technology that converts written text into spoken words. |
| VUI | 语音用户界面 | Interface allowing voice-based interaction with a device. |
12. Final Exam Tips for the Listening/Speaking Section | 口语/听力部分的终极考试技巧
On exam day, scan the paper for any questions related to sound, audio files, or speech technology. Jot down relevant formulas at the top of your working area to avoid confusion under pressure. For long-answer questions, note the command words: 'describe' requires more than a one-word answer; 'explain' demands reasons and steps. If asked to evaluate a voice-based system, always mention both advantages (hands-free, accessibility) and disadvantages (background noise, privacy concerns). Stay tuned into the language of the specification and allocate time proportionally to the marks.
考试当天，先浏览试卷中与声音、音频文件或语音技术相关的题目。在草稿区顶部快速记下相关公式，以免在紧张时混淆。对于简答题，注意指令词：“描述”要求的回答不能只写一个词；“解释”则需要给出原因和步骤。如果被要求评价一个基于语音的系统，务必提及优点（解放双手、无障碍）和缺点（背景噪音、隐私问题）。紧扣教学大纲的用语，并根据分值合理分配时间。
Published by TutorHao | Computer Science Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图