Helsinki Music ToolsSkip other details (including permanent urls, DOI, citation information)
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. Please contact email@example.com to use this work in a way not covered by the license. :
For more information, read Michigan Publishing's access and usage policy.
Page 352 ï~~Helsinki Music Tools Jukka-Pekka Kervinen (firstname.lastname@example.org) Kai Lassfolk (email@example.com) Department of Musicology P.O.Box 35 (Vironkatu 1) 00014 University of Helsinki Abstract Helsinki Music Tools (abbreviated as HMT) is a set of musical software tools. HMT works under the UNIX operating system. HMT consists of more than 50 programs that can be used for musical tasks such as algorithmic composition and music analysis. HMT programs behave like the standard UNIX tools and take advantage of standard UNIX features including input/output redirection. HMT provides no direct means of musical input or output. To compensate this, HMT can be used in combination with other music software such as a synthesis language. The current implementation supports Csound and the CCRMA Music Kit. 1 Introduction HMT consists of UNIX-like commands that are typically called from the shell. The HMT commands behave like their UNIX counterparts. They read and write ASCII-formatted data using the standard input/ output mechanism. Using ASCII allows to use standard UNIX commands in combination with HMT to perform musical tasks. The following example demonstrates how HMT commands are used and combined to generate and play a simple melody. $ hrnd 36 96 10 I ssgen I ssp Here the hrnd command prints 10 random numbers distributed uniformly between the minimum and maximum values of 36 and 96. hrnd prints the numbers to the standard output stream which is connected via a pipe to the next command, ssgen. ssgen reads standard input and generates an event list (called "score"). The score is in turn piped to ssp which plays it. hrnd is one of HMT's generative commands. There are nine commands for various stochastic distributions, two stochastic processes and an implementation of generative context-free grammars. These commands can be combined with the UNIX piping mechanism to form more complex random processes. ssgen is a tool for creating HMT scores. ssgen generates a "simple score". A simple score consists of timed events which in turn consist of parameter fields. The format resembles a conventional synthesis language such as Music V or Csound. An example simple score is shown below. vnr 1 0 1 60 30 vnr 1 1 1 62 60 vnr 1 2 1 64 90 vnr 1 3 1 65 120 vnr 1 4 2 67 150 The example score has five note events that form a melody line of c1, d', e1, fl and g'. Events are separated by newlines and parameter fields are separated by spaces and/or tabs. The first field in an event is an event type. The event type of a musical note is "vnr". The second field is an instrument number, an integer greater than or equal to 1. The third field is the starting time (called "attack time") of the event. It is represented in beats by a floating point number. The fourth field is duration of the event. The fifth field is a note number, an integer ranging from 0 to 127. The note numbers are the same as in MIDI. The sixth field is an amplitude value (or velocity), an integer ranging from 0 to 255. Attack times are calculated either from the beginning of the score or from start of the previous event. ssgen creates vnr events by reading note number values from standard input and filling the other fields with default values. This default action can be changed by using one of the following options: -i standard input is placed in the instrument number field -a standard input is placed in the attack time field -d standard input is placed in the duration field -v standard input is placed in the amplitude field ssgen can also be given a template score file as a command line argument. In the following example ssgen reads amplitude values from hind and fills in the other fields with values from vnr events stored in the file t. ssa. The output is redirected to u. ssa. $ hrnd 0 255 10 1 ssgen -v t.ssa >u.ssa 2P.05 352 ICMC Proceedings 1993
Page 353 ï~~HMT provides commands for score generation and manipulation. For example the following command prints attack times of a given score file. $ ssget -a testscore.ssa These values can be piped to a command that manipulates numerical data in various ways. ssget accepts the same options as ssgen. 2 Generative programs The generative programs give good examples of features common to most HMT commands. Generative programs allow arguments to be constants or variables. For example, the following command line $ hgsd 0 5 100 prints 100 normally distributed random numbers with mean value of 0 and variance of 5. The argument values 0 and 5 are treated as constants. In the following command $ hgsd a b 100 hgsd treats a and b as files containing numeric values. The program reads new mean and variance values from files a and b respectively for each random number it prints. The command terminates when 100 numbers are printed or if either a or b runs out of numbers. A variable argument may be either a valid UNIX filename or '-' which indicates standard input. Thus variable arguments may be piped from command to command as demonstrated by the following example. $ hgsd 0 5 100 1 hrwp 60 - 36 96 This command line combines the Gaussian distribution command hgsd and a random walk command hrwp. hrwp takes four arguments: start, jump, min and max. In the example hrwp uses jump values produced by hgsd. This generates a random process called Brownian motion [Dodge and Jerse, 1985]. By default all stochastic distributions and processes produce integer values. Also by default, the seed value for the random number generator is read from the system clock. These settings may be changed with the following options. - f Specify floating point output. This is not recognized by discrete distributions like Poisson. -s seed Specify a seed value to the random number generator. The last command line argument that specifies the amount of generated numbers is optional for stochastic distributions and processes. Omitting this argument prints an infinite random number series. Other stochastic commands include linear, triangular, Cauchy, Erlang and Weibull distributions and a simulated 1/f process by the Voss algorithm. 3 Scores and number sequences The two fundamental data representation formats used in HMT are scores and number sequences. A number sequence is a stream of consecutive numbers separated by spaces, tabs or newlines. Number sequences can represent any musical parameter, such as pitch, timing, amplitude or parameter specific to a synthesis program. A score represents a collection of musical parameters organized as events. An HMT score may contain any kind of data (not necessarily musical) as long as certain basic syntactic rules are followed. HMT simple scores follow strict rules and are limited in features. The idea behind simple scores is to provide an easy to learn starting point for newcomers and to provide a least common nominator for all scores. Simple scores can be converted to other formats such as MIDI files or synthesis languages (and back) with no loss of information. HMT provides tools for generating and manipulating simple scores. HMT also contains conversion programs between simple scores and other formats including CCRMA Music Kit scores [Boynton and Jaffe, 1989] and Csound [Vercoe, 1991]. HMT also allows the construction of more complex user-defined scores. It is however recommended that such a score be a superset of a simple score and should follow the same syntax. HMT has tools to create and modify arbitrary scores but no conversion programs to or from other music file formats are provided. The construction of conversion programs is left to the user. Score conversion from HMT to an other text file format such as a Csound score is however an easy task using e.g. the AWK programming language [Aho et al., 1988]. The command hgen generates user-defined score events by specifying default parameter field values with command line arguments. For example the following command line $ hrnd 36 96 5 I hgen vnr 1 0+1 1 - 128 prints a score similar to the example score presented in the first chapter. hgen accepts both constant and variable arguments as described in the previous chapter. The third argument ('0+1') has a notation specific to hgen. There '0' represents an attack time value for the first event. This is incremented by I for each following event hgen accepts any number of arguments. hget is a command for retrieving number sequences from score parameter fields. It works as ssget: (described in chapter 1) but has a different syntax to suit user-defined scores. ICMC Proceedings 1993 353 2P.05
Page 354 ï~~HMT contains commands for applying basic arithmetic operations on number sequences. These can be used for transposition or time scale modifications. Also inversion, linear interpolation and retrograde programs are provided. For example the following command $ ssget t2.ssa j hadd 12 1 ssgen t2.ssa prints a score where note numbers are transposed by an octave. 4 Context-free grammars Context-free grammars introduce a third data representation format as an addition to number sequences and scores. A grammar is stored in a text file. An example grammar is shown below. F ->ABC A -> 60 62 65 A -> B -> 72 68 67 C -> 55 52 36 C -> 55 57 60 48 A grammar file is processed by the command hgram that prints terminal symbols. hgram treats each line as a production that consists of a left-hand side (nonterminal), a production symbol ('->') and a righthand side (any number of terminals and/or non-terminals). If more than one production has the same lefthand side, one of them is picked at random. If a righthand side symbol does not exist as the left-hand side in any production, it is treated as terminal and thus printed if the appropriate production is chosen. Here is an example hgram call: $ hgram grammarfile where grammarfile contains a grammar. hgram accepts a random number generator seed value as described in chapter 2. 5 Other features HMT is designed to be easily modifiable by the user. In particular most data conversion programs are written in Bourne Shell and AWK to provide a starting point for modifications. Playback routines, such as the ssp command, allow users to write their own data conversion programs for the desired output format. ssp also allows to write a header file containing instrument definitions and other data specific to a synthesis language. HMT has been written mostly in the C programming language. Some commands, mainly conversion programs, were written in AWK and Bourne Shell. YACC and Lex were used in the generative grammar program. HMT should port easily to most UNIX System V and 4.3BSD implementations. The only major difficulty when porting HMT is the availability of musical I/O-routines. The CCRMA Music Kit for example is currently not portable beyond NeXT hardware. Also, UNIX lacks a standard MIDI application programming interface. 6 Summary HMT combined with standard UNIX tools provides a flexible and programmable music processing environment. Most HMT commands are general-purpose tools that generate or manipulate numerical data. These tools are not particularly "musical". What makes HMT a music software system is the score concept and'the score manipulation tools. HMT contains no sound synthesis or MIDI I/O routines. Instead it requires the use of other music software to perform these tasks. On NeXT computers HMT scores can be synthesized or converted to MIDI files using the CCRMA Music Kit. In other environments a synthesis language can be used. Perhaps most importantly HMT provides a framework for developing new composition and analysis programs. References [Aho et al., 1988] Alfred V. Aho, Brian W. Kernighan and Peter J. Weinberger. The AWK Programming Language. Addison-Wesley, Menlo Park, 1988. [Boynton and Jaffe, 1989] Lee Boynton and David A. Jaffe. An Overview of the Sound and Music Kits for the NeXT Computer. Computer Music Journal, Volume 13, Number 2, pp.48-55, 1989. [Dodge and Jerse, 1985] Charles Dodge and Thomas A. Jerse. Computer Music - Synthesis, Composition and Performance. Schirmer Books, New York, pp.289-290, 1985. [Vercoe, 1991] Barry L. Vercoe. Csound - A Manual for the Audio Processing System and Supporting Programs. Media Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts, 1991. 2P.05 354 ICMC Proceedings 1993