📚 Year 12 OCR Computer Science: Speaking & Listening Exam Prep | 口语/听力备考专项
For Year 12 OCR Computer Science students, ‘speaking and listening’ in the context of the specification refers not to a language oral exam, but to the fundamental concepts of sound representation, speech synthesis, and voice recognition. Understanding how computers capture, store and process human speech is a core part of the 1.2.3 units on data representation and 2.2.2 on software applications. This revision guide systematically breaks down these concepts, ensuring you can confidently tackle related exam questions.
对于 Year 12 OCR 计算机科学的学生来说，“口语与听力”并非指语言口试，而是指声音表示、语音合成及语音识别这些基础概念。理解计算机如何捕捉、存储和处理人类语音，是单元 1.2.3 数据表示和单元 2.2.2 软件应用的核心部分。本复习指南系统拆解这些概念，确保你能够自信应对相关考题。
1. Sound as an Analogue Signal | 声音的模拟信号本质
Sound is an analogue phenomenon, meaning it is continuously varying in both time and amplitude. To be processed by a digital computer, this smooth waveform must be converted into discrete digital data through a process known as analogue-to-digital conversion (ADC). A microphone captures sound waves and converts them into an analogue electrical voltage, which then must be sampled.
声音是一种模拟现象，意味着它在时间和振幅上都是连续变化的。要由数字计算机处理，这种平滑的波形必须通过模数转换（ADC）过程转换为离散的数字数据。麦克风捕捉声波并将其转换为模拟电信号，然后必须进行采样。
2. Sampling: Capturing the Waveform | 采样：抓取波形
Sampling is the process of measuring the amplitude of the sound wave at regular intervals. The sampling rate (or sampling frequency) is the number of samples taken per second, measured in Hertz (Hz). According to the Nyquist theorem, to accurately reproduce the original signal, the sampling rate must be at least twice the highest frequency present in the sound. For human speech (roughly 300 Hz – 3400 Hz), a sampling rate of 8 kHz is the baseline used in telephony, while CD-quality audio uses 44.1 kHz to cover the full 20 Hz – 20 kHz human hearing range.
采样是以固定间隔测量声波振幅的过程。采样率（或采样频率）是每秒采集的样本数量，以赫兹（Hz）为单位。根据奈奎斯特定理，要准确还原原始信号，采样率必须至少是声音中最高频率的两倍。对于人类语音（约 300 Hz – 3400 Hz），电话系统采用的基准采样率为 8 kHz，而 CD 品质音频使用 44.1 kHz 以覆盖人耳 20 Hz – 20 kHz 的可听范围。
3. Bit Depth and Quantisation | 位深度与量化
Bit depth (or sample resolution) determines the number of bits used to store each sample’s amplitude value. This process, called quantisation, rounds the measured amplitude to the nearest available integer level. A higher bit depth reduces quantisation error, improving the dynamic range and signal-to-noise ratio. For example, 8-bit audio provides only 256 discrete levels, which can sound grainy, whereas 16-bit audio gives 65,536 levels, delivering clear speech and music reproduction.
位深度（或采样分辨率）决定了每个采样点振幅值所用的比特数。这个过程称为量化，它将测得的振幅值四舍五入到最接近的可用整数级别。更高的位深度能减少量化误差，提升动态范围和信噪比。例如，8 位音频只提供 256 个离散级别，听起来可能带有颗粒感，而 16 位音频提供 65,536 个级别，可以呈现清晰的语音和音乐回放。
4. Calculating Audio File Size | 计算音频文件大小
The size of an uncompressed PCM audio file can be calculated using the formula: File size (bits) = sampling rate × bit depth × duration (seconds) × number of channels. For stereo (2 channels) speech recorded for 10 seconds at 8 kHz with 8-bit depth: File size = 8000 × 8 × 10 × 2 = 1,280,000 bits = 160,000 bytes = 156.25 KiB. Remember that OCR expects you to convert between bits, bytes, kilobytes (KiB), and megabytes (MiB) by dividing or multiplying by 8 and 1024 appropriately.
未压缩 PCM 音频文件的大小可以通过以下公式计算：文件大小（比特）= 采样率 × 位深度 × 时长（秒）× 声道数。对于以 8 kHz、8 位深度立体声录制 10 秒语音：文件大小 = 8000 × 8 × 10 × 2 = 1,280,000 比特 = 160,000 字节 = 156.25 KiB。注意，OCR 要求你能够通过在 8 和 1024 之间恰当乘除，完成比特、字节、千字节（KiB）和兆字节（MiB）之间的转换。
File size = Sampling rate (Hz) × Bit depth × Time (s) × Channels
5. MIDI vs Sampled Audio | MIDI 与采样音频对比
MIDI (Musical Instrument Digital Interface) is not a sound recording, but a set of instructions that tell a synthesiser how to produce music. Instead of storing sampled waveform data, a MIDI file stores event messages such as note on, note off, pitch, velocity, and instrument patch. This makes MIDI files extremely compact and editable, but they cannot represent speech—only synthesised sounds. Sampled audio captures real-world sound including speech, but results in much larger files.
MIDI（乐器数字接口）不是录音，而是一组指示合成器如何生成音乐的指令。MIDI 文件不存储采样波形数据，而是存储事件消息，如音符开、音符关、音高、力度和乐器音色。这使得 MIDI 文件极其紧凑且可编辑，但它们无法表示语音——只能表示合成的声音。采样音频则能捕捉包括语音在内的真实世界声音，但文件要大得多。
| Feature | Sampled Audio | MIDI |
|---|---|---|
| Stores | Sampled amplitudes | Performance instructions |
| Realism | Captures speech and natural sound | Only synthesised; no speech |
| File size | Large; depends on rate/depth | Very small |
6. Speech Synthesis: Computers that Speak | 语音合成：会说话的计算机
Speech synthesis is the artificial production of human speech. A text-to-speech (TTS) system converts written text into spoken words. This involves linguistic analysis to break down sentences, phoneme assignment, and concatenative synthesis where pre-recorded speech units (phonemes or diphones) are stitched together. Modern systems use deep learning to generate more natural intonation and rhythm, but the underlying principle remains: the computer crafts an audio output from stored sound fragments or parametric models.
语音合成是人工产生人类语音的过程。文本转语音（TTS）系统将书面文本转换为口语。这涉及语言学分析以分解句子、音素分配以及拼接合成，即将预录的语音单元（音素或双音素）拼接在一起。现代系统使用深度学习来生成更自然的语调和节奏，但基本原理不变：计算机利用存储的声音片段或参数模型来制作音频输出。
7. Speech Recognition: Computers that Listen | 语音识别：会听音的计算机
Speech recognition (or automatic speech recognition, ASR) enables a computer to identify and process human speech into text or commands. The audio input is first digitised, then features like mel-frequency cepstral coefficients (MFCCs) are extracted. Acoustic models match these features to phonemes, language models predict word sequences, and a decoder outputs the most likely transcription. Key challenges include background noise, accents, and homophones, which OCR exam questions often ask you to discuss.
语音识别（或自动语音识别，ASR）使计算机能够识别人类语音并将其处理为文本或命令。音频输入首先被数字化，然后提取如梅尔频率倒谱系数（MFCC）等特征。声学模型将这些特征与音素匹配，语言模型预测单词序列，解码器输出最可能的转录。主要挑战包括背景噪音、口音和同音异义词，OCR 考题经常要求你对此进行讨论。
8. The Role of Compression in Speech Transmission | 压缩在语音传输中的作用
Raw audio data is very large, so compression is vital for efficient storage and real-time transmission of speech. Lossless compression like FLAC preserves all original data and is used for archival. More common for speech are lossy codecs such as MP3 or the telephony-focused G.711, which exploit psychoacoustic principles to discard less audible information. Perceptual coding in lossy compression dramatically reduces bitrate while maintaining perceived quality, crucial for VoIP and streaming.
原始音频数据非常庞大，因此压缩对于语音的高效存储和实时传输至关重要。无损压缩如 FLAC 保留所有原始数据，用于存档。更常用于语音的是有损编解码器，如 MP3 或面向电话的 G.711，它们利用心理声学原理丢弃较不可闻的信息。有损压缩中的感知编码大幅降低比特率，同时保持感知质量，这对 VoIP 和流媒体至关重要。
9. Digital Audio Interfaces and Hardware | 数字音频接口与硬件
For a computer to both speak and listen, it requires appropriate audio hardware. Analogue-to-digital converters (ADCs) are needed at the input stage (microphone), and digital-to-analogue converters (DACs) at the output (speakers or headphones). The sound card or onboard audio codec handles these conversions. USB microphones and headsets integrate these circuits, simplifying setup. In the context of speech applications, the quality of these components directly impacts recognition accuracy and synthesis clarity.
计算机要能说能听，需要合适的音频硬件。输入端（麦克风）需要模数转换器（ADC），输出端（扬声器或耳机）需要数模转换器（DAC）。声卡或板载音频编解码器负责这些转换。USB 麦克风和耳机集成了这些电路，简化了设置。在语音应用中，这些组件的质量直接影响识别准确性和合成清晰度。
10. Exam Tips for Speaking & Listening Topics | 口语与听力主题的应试技巧
When facing OCR exam questions on sound representation and speech technology, follow these key strategies: (1) clearly distinguish between sampling rate and bit depth, explaining how each affects quality and file size; (2) always show your working for file-size calculations, with units at each step; (3) compare MIDI and sampled audio using a structured table-like answer; (4) link speech synthesis/recognition to the broader theme of computer applications and challenges; (5) use precise terminology such as quantisation, dynamic range, and concatenative synthesis. Past papers show that short-answer questions often ask for definitions, while longer questions require evaluation of trade-offs.
面对 OCR 关于声音表示和语音技术的考题，请遵循以下关键策略：（1）清楚区分采样率和位深度，解释各如何影响质量和文件大小；（2）在文件大小计算中始终展示步骤，每步都带上单位；（3）使用表格式的结构化答案对比 MIDI 和采样音频；（4）将语音合成/识别与更广泛的计算机应用及挑战主题联系起来；（5）使用精准术语，如量化、动态范围和拼接合成。历年试题显示，简答题常要求给出定义，而长问题则需要评估取舍。
Published by TutorHao | Computer Science Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导