版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
1、声音是一种由物体振动引发的物理现象Sound is a physical phenomenon produced by the vibration of matter, such as a violin string, or a block of wood. As the matter vibrates, pressure variations are created in the air surrounding it. This alteration of high and low pressure is propagated through the air in a wave-like m
2、otion. When a wave reaches the human ear, a sound is heard. 声音是一种由物体振动引发的物理现象,如小提琴的弦声等。物体的振动使其四周空气的压强产生变化,这种忽强忽弱变化以波的形式向四周传播,当被人耳所接收时,我们就听见了声音。 Sound / Audio 声音和音频技术涉及的主要是声波(声学信号)的处理。编码、录音机或数字音频带的存储、音乐及语音处理是这一领域的主要课题。 Sound methodology and audio techniques engage in processing these sound waves (aco
3、ustic signals). Important topics in this area are coding, storage on recorders or digital audio tapes, music and speed processing. 声音是由物体的振动产生的,这种振动引起了周围空气压强的振荡,我们称这种振荡的函数表现形式为波形.Sound is produced by the vibration of matter. During the vibration, pressure variations are created in the air surroundin
4、g it. The pattern of the oscillation is called a waveformBasic sound Concepts Figure 3.1:Oscillation of an air pressure wave.AmplitudeTimeAir PressureOne Period+如果每隔一定时间波形就重复相同的形状,那么就称这个时间为周期。一般然声音的波形既不会十分光滑也不会有很强的周期性。与非周期性的声音相比,具有周期性的声音听起来要悦耳得多。乐器的演奏、元音、鸟鸣及风声都是周期性的声音,而打击乐器声、咳嗽声、喷嚏声以及流水声则是非周期性的。The
5、waveform repeats the same shape at regular intervals and this portion is called a period. Since sound waves occur naturally, they are never perfectly smooth or uniformly periodic. However, sounds that display a recognizable periodicity tend to be more musical than those that are nonperiodic. Exles o
6、f periodic sound sources are musical instruments, vowel sounds, the whistling wind and bird songs. Nonperiodic sound sources include unpitched percussion instruments, coughs and sneezes and rushing water. 声音的频率是周期的倒数,它表示的是声音在1秒钟内的周期数,单位是赫兹(Hz)。千赫(kHz),即1000Hz,表示每秒振动1000次Boo87。声音按频率可作如下划分:The frequen
7、cy of a sound is the reciprocal value of the period; it represents the number periods in a second and is measured in hertz (Hz) or cycles per second (cps). A convenient abbreviation, kHz (kilohertz), is used to indicate thousands of oscillation per second: 1 kHz equals 1000 Hz Boo87. The frequency r
8、ange is divided into: Frequency 次声 020Hz 人耳能听见的声音 20Hz20KHz 超声 20KHz1GHz 特超声 1GHz10THz Infra-sound from 0 to 20 Hz Human hearing frequency range from 20Hz to 20kHzUltrasound from 20kHz to 1GHzHypersound from 1GHz to 10 THz 多媒体系统中使用的声音一般只限于人耳能听到的范围,我们把人耳所能听见的声音为音频(audio),将这个范围内的声波称为声学信号(acoustic sign
9、al)。例如语音是由人发出的声学信号,乐器的频率范围在20Hz之间。我们将语音和音乐之外的音频信号统称为噪音。Multimedia systems typically make use of sound only within the frequency range of human hearing. We will call sound within the human hearing range audio and the waves in this frequency range acoustic signals Boo87. For exle, speech is an acousti
10、c signal produced by humans; music signals have a frequency range between 20Hz and 20khz. Besides speech and music, we denote any other audio signal as noise. 声音有振幅,振幅的主观感觉是声音的大小。声音的振幅大小取决于空气压力波距平均值(也称平衡态)的最大偏移量。A sound also has an litude, a property subjectively heard as loudness. The litude of a s
11、ound is the measure of the displacement of the air pressure wave from its mean, or quiescent state. Amplitude计算机并不直接使用连续平滑的波形来表示声音,它是每隔固定的时间对波形的幅值进行采样,用得到的一系列数字量来表示声音。图3.2是经过数字采样的波形示意图。The smooth, continuous curve of a sound waveform is not directly represented in a computer. A computer measures the
12、 litude of the waveform at regular time interval to produce a series of numbers. Each of these measurements is a sle. Figure 3.2 illustrates one period of a digitally sled waveform. Computer Representation of Sound TimeSleSle height+0Figure 3.2:Sled waveform.使用模-数转换器(ADC)可以将音频信号转换成数字采样值,使用数-模转换器(DAC
13、)则可以将数字信号转换为模拟信号。AM79C30A芯片是一种ADC,SPARC工作站将其用于音频输入,而对于音频输出,桌面SPARC系统使用的是内置的扬声器。DAC也是一个标准的UNIX设备,例如SPARC服务器6xx系统没有自带的扬声器但支持外接的话筒和扬声器。 The Mechanism that converts and audio signal into digital sles is the Analog-to-Digital Converter (ADC). The reverse conversion is performed by a Digital-to-Analog Con
14、verter (DAC). The AM79C30A Digital Subscriber Controller chip is an exle of an ADC and is available on SPARCstationsTM. Desktop SPARCTMsystems include a built-in speaker for audio output. DAC is also available as a standard UNIXTMdevice. For exle, SPARCserver 6xx systems do not have an internal spea
15、ker, but support an external microphone and speaker. 采样频率表征的是单位时间内对波形采样44100次,这似乎超出了人耳所能听见声音的频率范围。但根据Nyguist采样定律,要从采样中完全恢复原始信号波形,采样频率必须至少是信号中最高频率的两倍,所以CD标准采样频率至少要是人耳所能听到频率上限20KHz的两倍,这样44100的采样率所能表示的频率上限为22050Hz。 The rate at which a continuous waveform (Figure 3.1) is sled is called the sling rate. L
16、ike frequencies, sling rates are measured in Hz. The CD standard sling rate of 44100 Hz means that the waveform is sled 44100 times per second. This seems to be above the frequency range the human ear can hear. However, the bandwidth (which in this case is 20000 Hz-20Hz=19980Hz) that digitally sled
17、audio signal can represent, is at most equal to half of the CD standard sling rate (44100Hz). This is an application of the Nyquist Sling theorem. (“For lossless digitization, the sling rate should be at least twice the maximum frequency responses.”) Hence, a sling rate of 44100Hz can only represent
18、 frequencies up to 22050Hz, a boundary much closer to that of human hearing.Sampling Rate采样是在离散的时间点上进行的,而采样值本身在计算机中也是离散的。采样值的精度取决于它用多少位来表示,这就是量化。例如8位量化可以表示256个不同值,而CD质量的16位量化可以表示65 536个值。图3.3是一个3位量化的示意图,可以看出3位量化只能表示8个值:0.75,0.5,0.25,0,0.25,0.5,0.75和1,因而量化位数越少,波形就越难辨认,还原后的声音质量也就越差(可能除了一片嗡嗡声之外什么都没有)Ju
19、st as a waveform is sled at discrete times, the value of the sle is also discrete. The resolution or quantization of a sle value depends on the number of bits used in measuring the height of the waveform. An 8-bit quantization yields 256 possible values; 16-bit CD-quality quantization results in ove
20、r 65536 values. Figure 3.3 presents a 3-bit quantization. The sled waveform with a 3-bit quantization results in only eight possible values: .75, .5 ,.25, 0, -.25, -.5, -.75 and 1. The shape of the waveform becomes less discernible with a lowered quantization, i.e., the lower the quantization, the l
21、ower the quality of the sound (the result might be a buzzing sound). QuantizationSle0.25.5.75-.25-.5-.75Sle HeightFigure 3.3:Three-bit quantization. 计算机必须有相应的输入输出设备才能进行声音信号的处理。话筒与内置扬声器分别是与ADC和DAC相连的音频输入与输出设备。Before sound can be processed, a computer needs input/output devices. Microphone jacks and b
22、uilt-in speakers are devices connected to an ADC and DAC, respectively for the input and output of audio. Sound Hardware 音频的两个重要格式参数是采样频率(如每秒8012.8次)和采样量化位数(如8位量化)。 The above exles of telephone-quality and CD-quality audio indicate that important format parameters for specification of audio are: sli
23、ng rate (e.g., 8012.8 sles/second) and sle quantization (e.g., 8-bit quantization). Audio Formats A MIDI interface has two different components:3.2.1 MIDI的基本概念 MIDI接口由两个不同的部分组成: Music一是与设备相连的硬件标准,它规定了乐器间的物理连接方式,要求乐器必须带有MIDI端口,它还对连接两个乐器的MIDI缆线和缆线上传输的电信号作了规定。Hardware connects the equipment. It specifi
24、es the physical connection between musical instruments, stipulates that a MIDI port is built into an instrument, specifies a Midi cable (which connects two instruments) and deals with electronic signals that are sent over the cable. 二是数据格式标准,它给出了硬件上传输的信息的编码方式。与音频数据格式( 3.1.2节)不同,MIDI数据格式不处理单个采样点的编码,它
25、包括乐器的定义、音符的开始与结束、基本频率及音量等描述。MIDI数据可以编码10个8度音程,也就是128个音符。A date format encodes the information traveling through the hardware. A MIDI data format does not include an encoding of individual sles as the audio format does (Section 3.1.2). Instead of individual sles, an instrument connected data format i
26、s used. The encoding includes, besides the instrument specification, the notion of the beginning and end of a note, basic frequency and sound volume. MIDI data allow an encoding of about 10 octaves, which corresponds to 128 notes. MIDI使用数字方式描述数据,数据被分成多组MIDI消息,一组MIDI消息代表一个音乐事件,这些音乐事件通常是演奏动作,比如击键、移动滑动
27、条、设置开关及调整脚踏板等 .The MIDI message communicates one musical event between machines. These musical events are usually actions that a musician performs while playing a musical instrument. The action might be pressing keys, moving slider controls, setting switches and adjusting foot pedals.比如弹奏钢琴,击键时由MIDI
28、接口产生一个MIDI消息,这个消息包含音符的开始及击键的强度等,它被传到另一台机器上,而当松开键时又产生一个相应的MIDI消息。10分钟的音乐,用MIDI来表示,只需200KB的MIDI数据,比CD质量音频编码少得多。When a musician presses a piano key, the MIDI interface creates a MIDI message where the beginning of the note with its stroke intensity is encoded. This message is transmitted to another mac
29、hine. In the moment the key is released, a corresponding signal (MIDI message) is transmitted again. For ten minutes of music, this process creates about 200 Kbytes of MIDI data, which is essentially less than the equivalent volume of a CD-audio coded stream in the same time.如果数字乐器满足了上述MIDI标准的两项要求,就
30、称之为MIDI设备(例如合成器),它就能通过MIDI通道与其他MIDI设备交换信息。MIDI标准规定可用16个通道来传输数据,每个通道对应一个MIDI设备(乐器)。由通道传输的音乐数据被接收端的合成器重新合成为音乐。MIDI还规定了128种乐器效果,其中包括噪音效果(如 、飞机等),每种乐器效果与一个惟一的数字相对应,例如0代表钢琴,12代表马林巴(marinba,一种乐器),40代表小提琴,73代表长笛等。If a musical instrument satisfies both components of the MIDI standard, the instrument is a MI
31、DI device (e.g., a synthesizer), capable of communicating with other MIDI devices through channels. The MIDI standard specifies 16 channels. A MIDI device (musical instrument ) is mapped to a channel. Music data, transmitted through a channel, are reproduced at the receiver side with the synthesizer
32、 instrument. The MIDI standard identifies 18 instruments, including noise effects (e.g., telephone, air craft), with unique numbers. For exle, 0 is for the Acoustic Grand Piano, 12 for the marimba, 40 for the violin, 73 for the flute, etc. 有些乐器一次只能演奏一具音符,如长笛,而有些乐器则可以同时演奏多个音符,如风琴。某个通道的合成器每次可演奏的最多音符个数
33、是合成器的一个主要参数。该参数一般在316之间。Some instruments allow only one note to be played at a time, such as the flute. Other instruments allow more than one note to be played simultaneously, such as the organ. The maximum number of simultaneously played notes per channel is a main property of each synthesizer. The
34、 range can be from 3 to 16 notes per channel. 要想将MIDI设备映射到通道上,必须设定设备的MIDI接收模式,共有4种接收模式: To tune a MIDI device to one or more channels, the device must be set to one of the MIDI reception modes. There are four modes:模式1 Omni On/Poly(所有 开/复音)模式2 Omni On/Mono(所有 开/单音)模式3 Omni Off/Poly(所有 关/复音)模式4 Omni
35、Off/Mono(所有 关/单音)Mode 1: Omni On/Poly; Mode 2: Omni On/Mono;Mode 3: Omni Off/Poly; Mode 4: Omni Off/Mono 通过MIDI接口,计算机可以控制各个乐器的输出,同时,通过这个接口,计算机还能接收、存储并处理经过编码的音乐数据。数据由键盘产生,可通过声音生成器还原为声音,用音序器存储数据,大多数情况下它还有编辑音乐数据的功能。在多媒体系统中,音序器是一个应用程序。Through the MIDI interface, a computer can control output of individu
36、al instruments, On the other hand, the computer can receive, store or process coded musical data through the same interface. The data are generated with a keyboard and reproduced through a sound generator. A sequencer can store data. Further, it may also modify the musical data. In a multimedia syst
37、em, the sequencer is a computer application. MIDI Devices 合成器是所有MIDI系统的核心。从外观上看合成器就像一个带有许多按钮的简单的钢琴键盘,但它决非如此简单。大部分合成器包括以下几个组成部分:The heart of any MIDI system is the MIDI synthesizer device. A typical synthesizer looks like a simple piano keyboard with a panel full of buttons, but it is far more (more
38、detailed information on synthesizers can be found in Boo87.). Most synthesizers have the following common components:声音生成器微处理器键盘控制面板辅助控制器存储器Sound GeneratorsMicroprocessorKeyboardControl PanelAuxiliary Controllers Memory 音序器是一个重要的MIDI设备,可以是击鼓机、计算机或专门设计的音序器。一开始音序器用来存储MIDI数据的、现在如果计算机充当音序器,还可以对音乐进行编辑。由于
39、数据是以数字形式表示的,所以很容易修改。音乐数据的表示方式有好几种,最常见的形式是用音符,这样音乐在计算机中以乐谱的形式出现,然后由音序器将乐谱中的音符转换为MIDI消息(见3.2.1和3.2.3节)。音乐数据的另一种表示形式是直接使用MIDI消息,但这要求用户研究每个通道需要的音乐事件类型及相应的时间依赖关系,而且MIDI消息的输入依赖于键盘的类型。An important MIDI device is a sequencer, which can be a drum machine, computer or dedicated sequencer. A sequencer was used
40、 originally as a storage server for generated MIDI data. Today, a sequencer, being a computer, becomes additionally a music editor. Data can be modified in a proper way because of their digital data representation. There are several possibilities to represent musical data. The most common representa
41、tion and manipulation of data are musical notes. The musical piece appears on the screen in the form of a sheet of music. The sequencer transforms the notes into MIDI messages (Sections 3.2.1, 3.2.3). Another representation is a direct input of MIDI messages. Here, the user specifies required musica
42、l events per channel with their time dependencies. This input depends on the keyboard type. MIDI消息所传送的既可以是数据信息也可以是确定音乐事件类型的控制信息。MIDI消息的格式包括状态字节和数据字节。状态字节总是MIDI消息的第一个字节,用来描述消息的类型,数据字节则是其余的部分。MIDI消息可分为两大类:MIDI messages transmit information between MIDI devices and determine what kinds of musical event
43、s can be passed from device to device. The format of MIDI messages consists of the status byte (the first byte of any MIDI message), which describes the kind of message, and data bytes (the following bytes). MIDI messages are divided into two different types:MIDI Messages通道消息只发向指定的设备,一般分为两种类型:Channe
44、l messages go only to specified devices. There are two types of channel message:Channel Messages 通道声音消息。它在MIDI设备间传送演奏的数据,包括键盘动作、控制动作及控制面板的变化。它用音调、振幅、音色及其他一些性质来描述音乐。在通道声音消息中,状态字节后至少有1个字节(通常2个字节)来描述声音的性质。诸如Note On,Note Off,Channel Pressure,Control Change等都属于通道声音消息。- Channel voice messages send actual
45、performance data between MIDI devices, describing keyboard action, controller action and control panel changes. They describe music by defining pitch, litude, timbre, duration and other sound qualities. Each message has at least one and usually two data bytes that accompany the status byte to descri
46、be these sound qualities. Exles of channel voice messages are Note On, Note Off, Channel Pressure, Control Change, etc.通道模式消息。用于确定MIDI接收设备对通道声音消息做出反应的方式,它可以设定MIDI设备的MIDI通道接收模式,可以中止错误音符的演奏,还可以影响对设备的局部控制。诸如Local Control,All Notes Off,Omni Mode Off等都是通道模式信息。 -Channel mode messages determine the way tha
47、t a receiving MIDI device responds to channel voice messages. They set the MIDI channel receiving modes for different MIDI devices, stop spurious notes from playing and affect local control of a device. Exles of such messages are Local Control, All Notes Off, Omni Off, etc.因为没有指定通道号,所以系统消息发向MIDI系统中的
48、所有设备。系统消息共分为三类:System messages go to all devices in a MIDI system because no channel numbers are specified. There are three types of system messages:System Messages 实时系统消息。这是一个只有1字节的简短消息,它能携带特殊的数据。这些消息使MIDI设备在演奏中保持同步,因此一旦需要类信息就应立即发送。为了避免延尺,如果需要,这些消息可以夹在其他消息的中间发送。System Reset,Timing Clock(MIDI时钟)等都是这
49、类消息。-system real-time messages are very short and simple, consisting of only one byte. They carry extra data with them. These messages synchronize the timing of MIDI devices in performance; therefore, it is important that they be sent at precisely the time they are required. To avoid delays, these m
50、essages are sent in the middle of other messages, if necessary. Exles of such messages are System Reset, Timing Clock (MIDI clock), etc一般系统消息。一般系统消息可以使音序器和合成器做好演奏准备。这些消息可以选择乐曲确定演奏起点,如果需要,还可以对合成器调音。Song Select,Tune Request 都是一般系统消息。-system common messages are commands that prepare sequencers and synt
51、hesizers to play a song. The various messages enable you to select a song, find a common starting place in the song and tune all the synthesizers if they need tuning. Exles are Song Select, Tune Request, etc.专用系统消息。专门系统消息使得MIDI制造商可以在各自的产品上使用自定义的MIDI消息,这些消息以系统专用消息开头,通过它确定生产厂家,最后又以系统专用消息结尾。System excl
52、usive messages allow MIDI manufacturers to create customized MIDI messages to send between their MIDI devices. This coding starts with a system-exclusive-message, where the manufacturer is specified, and end-of-exclusive message.人和机器都可以感知、理解和产生语音。人能很好地适应说话者的语音差另和语言习惯,因此即使存在方言和发音差别,人还是能正确地理解会话。人脑能很好地
53、区分语音和噪音。但要做到这几点最好使用两只耳朵,因为只用一只耳朵,滤波会困难很多。Speech can be “perceived,” “understood” and “generated” by humans and also by machines. A human adjusts himself/herself very efficiently to different speakers and their speech habits. Despite different dialects and pronunciation, the speech can be will unders
54、tood by humans. The brain can recognize the very fine line between speech and noise. For this purpose, both ears are used, because filtering with only one ear is substantially more difficult for the listener.Speech人的语言信号包括一个被称为音调的主观低频分量,它不与频率成比例。人耳对600Hz到6 000Hz波段的声音最为敏感,Fletscher和Munson指出,人耳对1KHz左右
55、的声音最为敏感,而对高频声音的敏感性则大大降低。在语言处理中常用到的语音特性有两个:The human speech signal comprises a subjective lowest spectral component known as the pitch, which is not proportional to frequency. The human ear is most sensitive in the range from 600Hz to 6000Hz. Fletscher and Munson have shown that the human ear is subs
56、tantially less sensitive to low and very high frequencies than to frequencies around 1 kHz. Speech signals have two properties which can be used in speech processing: 语音信号在一定时间段内有近似周期性的行为,因此,可以把这些信号看做是30ms的半静止(quasi-stationary)信号。Voiced speech signals show during certain time intervals almost period
57、ic behavior. Therefore, we can consider these signals as quasi-stationary signals for around 30 milliseconds.语音信号的频谱有35个频率段的共振峰,这些称为共振峰的最大值是由声带共振引起的。The spectrum of audio signals shows characteristic maxima, which are mostly 3-5 frequency bands. These maxima, called formants, occur because of resona
58、nces of the vocal tract. 语音生成的研究有着很深的渊源,早在19世纪中叶,Helmholtz就使用成对的机械共鸣器组成机械志带来产生语音。到1940年,Dudley用电磁振动代替机械振动,试制成功了第一个语言合成器。Speech generation research has a long history. By the middle of the 19th century, Helmholtz had already built a mechanical vocal tract coupling together several mechanical resonato
59、rs with which sound could be generated. In 1940, Dudley produced the first speech synthesizer through imitation of mechanical vibration using electrical oscillation.Speech Generation 语音生成的一项关键技术是信号的实时生成,如果满足了这一要求,语音输出系统就可以不经预处理直接将文本转换为语音。有些软件要求的词汇量不大,比如 报时,但大部分应用要求有相当大的词汇量。An important requirement f
60、or speech generation is real-time signal generation. With such a requirement met, a output system could transform text into speech automatically without any lengthy preprocessing. Some applications only need a limited vocabulary; However, most applications need a large vocabulary, if not an unlimite
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 2026年市场监管系统事业单位真题及答案
- 2026年安徽省北师大版五年级数学下册单元重难点检测试卷
- 2026年气道管理与气管插管试题(含答案)
- 2026年建筑施工安全生产岗位安全应急效率提升考试题库及答案
- 2025年吴忠市红寺堡区辅警协警招聘考试试题及答案详解
- 2026年法律职业资格考试(客观题)真题汇编及答案
- 2025年通信设备工程师初级职业鉴定考试题库及答案
- 2025年上半年教师资格证考试中学《综合素质》真题及答案解析
- 2025年全国通信专业技术人员职业水平考试试卷及答案
- 2026注册会计师经济法试题库及答案
- 2025年旅游管理运营能力考核试题及答案解析
- 血透患者运动康复指导
- 幼儿园教师意识形态培训内容
- 肥料成品库管理制度
- 公司抵质押品管理制度
- T/QQCA 003-2022藏医坛轮(札麦承廓)疗法技术规范
- 大额存单认购协议书模板
- 2025年办公园区前期物业管理服务协议
- 2025年广东省惠州市惠城区市场监督管理局招聘历年高频重点提升(共500题)附带答案详解
- GB/T 45085-2024再生资源回收利用网络信息存证规范
- 服饰品配件设计概论
评论
0/150
提交评论