Music – Instruments – Electrical musical tone generation
Reexamination Certificate
1999-04-21
2001-03-13
Nappi, Robert E. (Department: 2837)
Music
Instruments
Electrical musical tone generation
C084S615000, C084S649000, C084S653000, C434S30700R
Reexamination Certificate
active
06201176
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to the field of music systems and, in particular, to the identification and retrieval of particular pieces of music or alternately, attributes of a desired piece of music, from a music database on the basis of a query composed of desired features and conditional statements.
BACKGROUND OF THE INVENTION
Retrieval of music or music attributes from a database requires, in common with generic database functionality, a query method which is powerful and flexible, and preferably intuitively meaningful to the user. This in turn requires that the database contain music which has been classified in a manner which is conducive to systematic search and sort procedures. This latter aspect in turn requires that pieces of music be characterized in a manner which permits of such classification.
Thus a hierarchy of requirements or elements which make up a music database system are as follows:
characterizing music using attributes useful in a classification scheme,
classifying music in a meaningful searchable structure, and
querying the database so formed, to yield meaningful results.
The hierarchy has been defined “bottom up” since this presents a more meaningful progression by which the invention can be described.
When considering audio signals in general, and in particular those relating to music, the nature of the signals may be considered in terms of various attributes which are intuitively meaningful. These attributes include, among others, tempo, loudness, pitch and timbre. Timbre can be considered to be made up of a number of constituent sub-features including “sharpness” and “percussivity”. These features can be extracted from music and are useful in characterizing the music for a classification scheme.
The publication entitled “Using Bandpass and Comb Filters to Beat-track Digital Audio” by Eric D. Scheirer (MIT Media Laboratory, Dec. 20, 1996) discloses a method for extraction of rhythm information or “beat track” from digital audio representing music. An “amplitude-modulated noise” signal is produced by processing a musical signal through a filter bank of bandpass filters. A similar operation is also performed on a white noise signal from a pseudo-random generator. Subsequently the amplitude of each band of the noise signal is modulated with the amplitude envelope of the corresponding band of the musical filter bank output. Finally the resulting amplitude modulated noise signals are summed together to form an output signal. It is claimed that the resulting noise signal has a rhythmic percept which is significantly the same as that of the original music signal. The method described can run in real-time on a very fast desktop workstation or alternately, a multiprocessor architecture may be utilized. This method suffers from the disadvantage of being highly computationally intensive.
Percussivity is that attribute which relates to a family of musical instruments known as “percussion” when considering an orchestra or a band. This family includes such musical instruments as drums, cymbals, castanets and others. Processing of audio signals in general and musical signals in particular, benefits from the ability to estimate various attributes of the signals, and the present invention is concerned with estimating the attribute of percussivity.
A number of different methods have been used to estimate percussivity of a given signal, such methods including those broadly based upon:
short-time power analysis,
statistical analysis of signal amplitude, and
comparison of harmonic spectral component power with total spectral power.
Short-time signal power estimation involves calculation of an equivalent power (or an approximation thereof) within a short segment or “window” of a signal under consideration. The power estimate can be compared to a threshold in order to determine whether the portion of the signal within the window is percussive in nature. Alternatively, the power estimate can be compared to a sliding scale of thresholds, and the percussive content of the signal classified with reference to the range of thresholds.
Statistical analysis of signal amplitude is typically based upon a “running mean” or average signal amplitude value, where the mean is determined for a window which slides across the signal under consideration. By sliding the window, the running mean is determined over a pre-determined time period of interest. The mean value at each window position is compared to mean values for other windows in a neighborhood in order to determine whether signal variations in the running mean are sufficiently large to signify that the signal is percussive.
Harmonic spectral component power analysis involves taking a windowed Fourier transform of the signal in question over the time period of interest, and then examining the resulting set of spectral components. The spectral components which are indicative of harmonic series are removed. It is noted that such harmonic series components typically represent local maxima in the overall spectral envelope of the signal. After removing the harmonic series spectral components, remaining spectral components substantially consist only of the inharmonic components of the signal, these being considered to represent percussive components of the signal. The total power in these inharmonic components is determined and compared with a total signal power for all components, harmonic and non-harmonic, to yield an indication of percussivity.
The aforementioned analysis methods are typically intended to identify a range of signal attributes, and thus suffer from relatively limited accuracy, and a tendency to produce false or unreliable percussivity estimates. The methods are also relatively complex and thus expensive to implement, particularly the harmonic spectral component estimation method.
U.S. Pat. No. 5,616,876 (Cluts et al) entitled “System and Methods for Selecting Music on the Basis of Subjective Content” describes an interactive network providing music to subscribers which allows a subscriber to use a seed song to identify other songs similar to the seed song, the similarity between songs being based on the subjective content of the songs, as reflected in style tables prepared by editors. The system and methods described in this publication are based on the manual categorization of music, with the attendant requirement for human participation in the process, with the resultant speed, accuracy and repeatability of the process limited by human attributes.
The publication entitled “Content-Based Classification, Search, and Retrieval of Audio” by Erling ct al (IEEE Multimedia Vol. 3, No. 3, 1996, pp. 27-36) discloses indexing and retrieving short audio files (i.e. “sounds”) from a database. Features from the sound in question are extracted, and feature vectors based on statistical measures relating to the features are generated. Both the sound and the set of feature vectors are stored in a database for later search and retrieval. A method of feature comparison is used to determine whether or not a selected sound is similar to another sound stored in the database. The feature set selected does not include tempo and thus the system will not perform well in differentiating between pieces of music. Furthermore, the method determines features which provide scalar statistical measures over short time windows. Furthermore, the method uses features such as bandwidth which are not readily conceptualized in terms of impact of music selection.
It is seen from the above that existing arrangements have shortcomings in all elements in the hierarchy of requirements described, and it is an object of the invention to ameliorate one or more disadvantages of the prior art.
SUMMARY OF THE INVENTION
According to one aspect of the invention, there is provided a method for querying a music database, which contains a plurality of pieces of music. The method involves classifying the pieces of the music in the database, using feature extraction. The classification involves segmenting each piece of music into a plurality of windo
Canon Kabushiki Kaisha
Fitzpatrick ,Cella, Harper & Scinto
Fletcher Marlon T.
Nappi Robert E.
LandOfFree
System and method for querying a music database does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for querying a music database, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for querying a music database will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2508110