Poisson point process modeling for polyphonic music transcription

Peaks detected in the frequency domain spectrum of a musical chord are modeled as realizations of a nonhomogeneous Poisson point process. When several notes are superimposed to make a chord, the processes for individual notes combine to give another Poisson process, whose likelihood is easily computable. This avoids a data association step linking individual harmonics explicitly with detected peaks in the spectrum. The likelihood function is ideal for Bayesian inference about the unknown note frequencies in a chord. Here, maximum likelihood estimation of fundamental frequencies shows very promising performance on real polyphonic piano music recordings.     

Musical variations from a chaotic mapping

A chaotic mapping provides a technique for generating musical variations of an original work. This technique, based on the sensitivity of chaotic trajectories to initial conditions, produces changes in the pitch sequence of a piece. A sequence of musical pitches {p(i)}, i.e., any piece ranging from Bach (or earlier) to contemporary music, is paired with the x-components {x(i)} of a Lorenz chaotic trajectory. Each p(i) is marked on the x axis at the point designated by its x(i). In this way, the x axis becomes a pitch axis configured according to the notes of the original composition. Then, a second chaotic trajectory, whose initial condition differs from the first, is launched. Its x-components trigger pitches on the pitch axis (via the mapping) that vary in sequence from the original work, thus creating a variation. There are virtually an unlimited number of variations possible, many appealing to expert and nonexpert alike. (c) 1996 American Institute of Physics.     

一个快速自动音乐记谱方法

自动音乐记谱是音乐信号处理中的关键技术。本文描述了一个快速的自动复音音乐记谱方法。该方法采用回声器时频分析(RTFI)作为时频分析工具,主要由两个阶段组成,能量基的音符切分和多基频估计。本文所采用的多基频估计方法首先将RTFI能量谱按照谐音组合原理转换为基频能量谱,并基于基频能量谱采用简单的峰拾起方法对基频做初步估计;然后根据频谱不规律性和乐音谐音结构的基本假定,消除初步估计中的错误预测。     

Monaural room acoustic parameters from music and speech

This paper compares two methods for extracting room acoustic parameters from reverberated speech and music. An approach which uses statistical machine learning, previously developed for speech, is extended to work with music. For speech, reverberation time estimations are within a perceptual difference limen of the true value. For music, virtually all early decay time estimations are within a difference limen of the true value. The estimation accuracy is not good enough in other cases due to differences between the simulated data set used to develop the empirical model and real rooms. The second method carries out a maximum likelihood estimation on decay phases at the end of notes or speech utterances. This paper extends the method to estimate parameters relating to the balance of early and late energies in the impulse response. For reverberation time and speech, the method provides estimations which are within the perceptual difference limen of the true value. For other parameters such as clarity, the estimations are not sufficiently accurate due to the natural reverberance of the excitation signals. Speech is a better test signal than music because of the greater periods of silence in the signal, although music is needed for low frequency measurement.     

Computational models of music perception and cognition I: The perceptual and cognitive processing chain

We present a review on perception and cognition models designed for or applicable to music. An emphasis is put on computational implementations. We include findings from different disciplines: neuroscience, psychology, cognitive science, artificial intelligence, and musicology. The article summarizes the methodology that these disciplines use to approach the phenomena of music understanding, the localization of musical processes in the brain, and the flow of cognitive operations involved in turning physical signals into musical symbols, going from the transducers to the memory systems of the brain. We discuss formal models developed to emulate, explain and predict phenomena involved in early auditory processing, pitch processing, grouping, source separation, and music structure computation. We cover generic computational architectures of attention, memory, and expectation that can be instantiated and tuned to deal with specific musical phenomena. Criteria for the evaluation of such models are presented and discussed. Thereby, we lay out the general framework that provides the basis for the discussion of domain-specific music models in Part II.     

一种音乐情绪参数化的方法

针对目前基于情绪的音乐分类研究存在的弊端,为了方便音乐检索,本文提出一种音乐情绪参数化的方法。该方法通过提取反映音乐情绪的特征向量,然后利用fisher算法进行维数压缩,再通过大量的音乐样本训练得到节奏、音调和音色3个描述音乐情绪的参数,参数的大小反映了情绪的强弱。实验结果表明,音乐情绪参数化的结果符合音乐实际的情绪。     

A theoretical study of the vibration and acoustics of ancient Chinese bells

In this paper, the acoustics of an ancient Chinese bell, which was made some 3000 years B.C., is studied theoretically. In ancient times, a set of the bells was used as a musical instrument. Unlike a western church bell and an ancient Indian bell, an ancient Chinese bell has two interesting acoustics. First, two tones can be heard separately as the bell is struck at two special points. The interval between the two pitches is always a minor or major third. Second, tones of the bell attenuate quickly, which is necessary for a musical instrument. So, an ancient Chinese bell is sometimes called a two-tone bell or a music bell. Although a three-dimensional model should be used to simulate the acoustics of the bell, a simplified model proposed in this paper does give some insight. Based on the lens-shaped cross section of an ancient Chinese bell, two tones of an ancient Chinese bell can be simulated by the vibration of a double-circular arch and the quick attenuation of tones can be simulated by acoustics of a cylinder with the lens-shaped cross section like a double-circular arch. Numerical results on the vibration and acoustics of the models are presented.     

椰胡声功率级的半消声室测试

椰胡是具有地方特色的拉弦乐器之一,多用于演奏广东音乐和广东戏曲、曲艺的伴奏。但对其发声强度一直未进行过科学的测试。本文介绍在华南理工大学半消声室内参照ISO(GB)标准对椰胡声功率级的测量工作。由两位资深乐师分别用两把椰胡在pp,mp,f和ff力度下演奏单音、音阶和乐曲,对每一把椰胡的每一测试内容,由十通道测试设备同步测试中心频率为100～10000 Hz的1/3倍频带声压级谱,通过计算获得每把椰胡在演奏每项内容时的声功率级和动态范围。通过对两把椰胡的测试结果进行平均,获得该乐器在演奏上述内容时的典型声功率级数值及频谱。文中并将半消声室内的测试结果与混响室内的测试结果相对比,探讨测试环境对测试结果的影响。民族乐器发声强度及其频谱特性的获得是开展民族音乐厅堂音质研究的基础。     

Dial A440 for absolute pitch: absolute pitch memory by non-absolute pitch possessors

Listeners without absolute (or "perfect") pitch have difficulty identifying or producing isolated musical pitches from memory. Instead, they process the relative pattern of pitches, which remains invariant across pitch transposition. Musically untrained non-absolute pitch possessors demonstrated absolute pitch memory for the telephone dial tone, a stimulus that is always heard at the same absolute frequency. Listeners accurately classified pitch-shifted versions of the dial tone as "normal," "higher than normal" or "lower than normal." However, the role of relative pitch processing was also evident, in that listeners' pitch judgments were also sensitive to the frequency range of stimuli.     

Pitch discrimination abilities in classical Arab-music listeners

The current experiment examined pitch discrimination thresholds in listeners of classical Arab music, and listeners of Western popular music. Classical Arab music is characterized by modes (“Maqamat”, plural of “Maqam” in Arabic language) of which the smallest interval is a quarter tone. In contrast, the smallest interval in Western music is a semitone. We hypothesized that daily exposure to a musical style involving minuscule pitch differences may have a positive effect on pitch discrimination abilities. Results demonstrate superior pitch discrimination abilities in the classical Arab music listeners. These results indicate that musical cultures may differ in their influence on perceptual abilities, depending on their basic acoustic characteristics.     

Fine-grained pitch processing of music and speech in congenital amusia

Congenital amusia is a lifelong disorder of music processing that has been ascribed to impaired pitch perception and memory. The present study tested a large group of amusics (n=17) and provided evidence that their pitch deficit affects pitch processing in speech to a lesser extent: Fine-grained pitch discrimination was better in spoken syllables than in acoustically matched tones. Unlike amusics, control participants performed fine-grained pitch discrimination better for musical material than for verbal material. These findings suggest that pitch extraction can be influenced by the nature of the material (music vs speech), and that amusics' pitch deficit is not restricted to musical material, but extends to segmented speech events.     

Absolute pitch among American and Chinese conservatory students: prevalence differences, and evidence for a speech-related critical period

Absolute pitch is extremely rare in the U.S. and Europe; this rarity has so far been unexplained. This paper reports a substantial difference in the prevalence of absolute pitch in two normal populations, in a large-scale study employing an on-site test, without self-selection from within the target populations. Music conservatory students in the U.S. and China were tested. The Chinese subjects spoke the tone language Mandarin, in which pitch is involved in conveying the meaning of words. The American subjects were nontone language speakers. The earlier the age of onset of musical training, the greater the prevalence of absolute pitch; however, its prevalence was far greater among the Chinese than the U.S. students for each level of age of onset of musical training. The findings suggest that the potential for acquiring absolute pitch may be universal, and may be realized by enabling infants to associate pitches with verbal labels during the critical period for acquisition of features of their native language.     

Perception of acoustic scale and size in musical instrument sounds

There is size information in natural sounds. For example, as humans grow in height, their vocal tracts increase in length, producing a predictable decrease in the formant frequencies of speech sounds. Recent studies have shown that listeners can make fine discriminations about which of two speakers has the longer vocal tract, supporting the view that the auditory system discriminates changes on the acoustic-scale dimension. Listeners can also recognize vowels scaled well beyond the range of vocal tracts normally experienced, indicating that perception is robust to changes in acoustic scale. This paper reports two perceptual experiments designed to extend research on acoustic scale and size perception to the domain of musical sounds: The first study shows that listeners can discriminate the scale of musical instrument sounds reliably, although not quite as well as for voices. The second experiment shows that listeners can recognize the family of an instrument sound which has been modified in pitch and scale beyond the range of normal experience. We conclude that processing of acoustic scale in music perception is very similar to processing of acoustic scale in speech perception.     

Perception of musical and lexical tones by Taiwanese-speaking musicians

This study explored the relationship between music and speech by examining absolute pitch and lexical tone perception. Taiwanese-speaking musicians were asked to identify musical tones without a reference pitch and multispeaker Taiwanese level tones without acoustic cues typically present for speaker normalization. The results showed that a high percentage of the participants (65% with an exact match required and 81% with one-semitone errors allowed) possessed absolute pitch, as measured by the musical tone identification task. A negative correlation was found between occurrence of absolute pitch and age of onset of musical training, suggesting that the acquisition of absolute pitch resembles the acquisition of speech. The participants were able to identify multispeaker Taiwanese level tones with above-chance accuracy, even though the acoustic cues typically present for speaker normalization were not available in the stimuli. No correlations were found between the performance in musical tone identification and the performance in Taiwanese tone identification. Potential reasons for the lack of association between the two tasks are discussed.     

基于新型神经网络的乐曲内容检索技术研究

基于文字符号的检索体系,对于音乐材料效果不佳,基于内容的音乐信息检索会成为今后发展的关键。针对音乐检索技术的核心——音频特征描述和特征匹配,本文以音高和相对节奏为音乐作品的关键特征,发展了非毗邻层连接的前馈神经网络结构,给出了误差反传训练算法的分类器,并进行了试验研究。结果表明,非毗邻层连接的前馈神经网络结构有优越的识别性能和极快的收敛速度。     

中国筝的声功率级测试

筝是中国古老的弹弦乐器。但迄今为止,对其声功率级一直未进行科学的测定。本文在一混响室内根据ISO及GB标准,对二十一弦筝的声功率级进行了首次测定。两位资深乐师在混响室内分别演奏各自的乐器,通过围绕乐师和乐器布置的四通道测试设备,对筝所辐射的声功率级和动态范围进行测定。测试结果表明,中国筝在以不同力度演奏单音、音阶和乐曲时所辐射的声功率级及其频率特性均有所不同。考虑到乐器演奏音阶时所辐射的声功率级及其动态范围与演奏乐曲时的声功率级接近,并且,音乐的空间感也大都在乐器以f力度演奏乐曲的强音标志乐段时最为显著,故此我们建议中国筝所辐射的声功率级用其以f力度演奏音阶时的平均声功率级表示。本文测试的两架中国筝以f力度演奏音阶时的平均声功率级为85.9dB。文中不仅首次公布了中国筝声功率级的测试结果,并且所介绍的测试方法对其它乐器声功率级测试也具有借鉴意义。民族乐器所辐射的声音性能的确定是民族音乐厅堂音质研究的基础。     

语后聋人工耳蜗使用者电刺激听觉部位音调感知研究

系统地研究了人工耳蜗植入者的电刺激听觉部位音调感知,全面地探讨了部位音调感知与人工耳蜗植入者言语识别和音乐感知的关系。4位成人语后聋人工耳蜗植入者参与了该研究。通过电极音调排序测试度量植入者的部位音调感知能力。言语能力测试和音乐音高分辨测试分别用米考察植入者的言语识别和音乐感知能力。结果显示,随着电极刺激部位从蜗尖移向蜗底,所有受试者均可获得从"低"到"高"的音调感知变化,但个体差异较大。受试者的言语识别结果与其电刺激听觉部位音调感知能力相关,但受到天花板效应影响,对应关系并不明显。受试者的音乐音高分辨成绩与其电刺激听觉部位音调感知能力呈较好的对应关系。结果表明,当前人工耳蜗声音编码策略所传递的声信号特征已可使植入者获得良好的言语识别效果;且安静环境下言语识别对植入者的部位音调感知能力要求不高。但当前的声音编码策略并未能有效对音乐信号进行编码;植入者在理解音乐这类复杂声信号时,其电刺激听觉部位音调感知能力一定程度决定了其听音效果。