Music Improvisation with Artificial Intelligence
[ April 2019 ]
I have started a project to generate music using Artificial Intelligence, inspired by the work by Google Magenta. The project will make use of the Magenta Resources, but will also be original in that it will be a new combination of approaches with a specific musical context in mind.https://magenta.tensorflow.org/
The Goal is to design a music generator that is capable of generating melodies that are both musically expressive and have a certain coherency. An interesting approach to this is to first train an LSTM on a dataset consisting of composed melodies, then tune this generator using rule based Q-learning (Jaques,Gu, Turner & Eck, 2017)
A goal in either the first or second phase of the project is also that the music generator is going to also be somewhat responsive to another player. This aspect will draw inspiration from the work done by Google Creative Lab on AI Duet.https://magenta.tensorflow.org/2017/02/16/ai-duet
Jaques, N., Gu,S., Turner, R E., & Eck, D. 2017.'TUNING RECURRENT NEURAL NETWORKS WITH REINFORCEMENT LEARNING'. NIPS 2016 Deep Reinforcement Learning Workshop. Available at: https://arxiv.org/pdf/1611.02796v2.pdf[Accessed 04.04.2019]
CAS Trumpet Mod Prototype
[ January 2014]
CAS Trumpet Mod Prototype
Last semester, I built a prototype of the CAS Trumpet Mod. This is a meta-trumpet; an augmented instrument system consisting of attachable sensors, sound engine and mapping strategies enabling simultaneous interaction with both the trumpet and different software based musical tools.
The project was both the semester assignment for the class MUS2830 Interactive Music at the University of Oslo, and the first part of the Chasm Interactive Music Technology Project.
Although the sketches and thought processes leading up to the development of the CAS Trumpet Mod has been largely characterized by on-and-off simultaneous brainstorming around different challenges and ideas, I have structured the information to provide a good overview of the different topic areas. The presentation is based around the approach to designing Digital Musical Instruments (DMI´s) suggested by Miranda and Wanderley in New Digital Musical Instruments: Control and Interaction Beyond the Key (Miranda/Wanderley 2006:4).
Anyone who wish to do so may emulate the work presented in this paper as long as the development is for non-commercial use, does not violate the rights of other research material presented in this paper, and all information presented in this paper is referred according to “god skikk”.
This report therefore both on developing the system and on various challenges related to its musical interactivity.
For the CAS Trumpet Mod Prototype I have used sensors from the PhigdetInterfaceKit. The digital platform used is MAX by Cycling 74.
Background As mentioned, the development of the CAS Trumpet Mod is a part of the Chasm Interactive Technology Project. Through this project I am seeking new ways of creating music. “I aim to create flexible and expressive electro-acoustic setups enabling the performer to instantly create any soundscape that comes to mind”(Svalesen). The purpose of the project is to develop customized tools for me to use both when I play live improvised electro-acoustic soundscapes and when I wish to enhance/change the role of the trumpet in more conventional bands.
This is the set up I currently use for improvising soundscapes.
It consists of a Kaosspad, a Zvex Fuzzfactory, a Kaossilator, a trumpet and a microphone. I usually play into the microphone, which sends the sound directly into the Kaosspad. Here, the sound is processed with effects, sampled, re-sampled with new effects etc. The Kaossilator is essentially a small touch synthesizer with 100 different pre-programmed sounds. The signal from this is sent through the Zvex Fuzzfactory, which is a stompbox fuzz, and into the Kaosspad. The Kaosspad cannot get input from the Kaossilator (line input) and Microphone (Mic input) simultaneously and thus operating the Kaosspad while playing Trumpet has quite a few restrictions. For example, to quickly change pitch on the trumpet while maintaining a stable tone requires both hands, making it difficult to play faster phrases while interacting with the electronics. When using my current set up, I usually alternate between being either mostly an electronics player or mostly a trumpet player.
To solve this problem, I have been working on a concept on how to integrate the electronics with my trumpet playing. The goal is to create a device that enables the electronics and the trumpet to be experienced by the performer as one combined instrument, rather than two separate.
Another limit to my current set up is the time it takes to heavily process sounds. I first have to sample the sound, and then resample for each effect I wish to add. Although this works well when one builds slowly evolving soundscapes, it is far from ideal when more dynamic and rapidly evolving sounds are required. As a part of my project I am therefore also stepping into the world of computer based live electronics, planning to use Max and Ableton Live as platforms for my new controller. Through these platforms, I wish to be able to create a nearly limitless setup, so that any soundscape that may come to mind can be unfolded in a matter of seconds.
With this in mind, I decided upon the following goal for my assignment:
“To develop a functioning prototype which enables simultaneous interaction with both trumpet and electronics, while exploring the different challenges related to musical interaction with this controller”
Planning and development:
I wanted to be able to utilize the enhanced possibilities the controller could provide and combine this with the previously well-learned trumpet technique. I therefore chose to derive the signal data for the digital part of the controller from additional physical gestures available in normal interaction with the trumpet. The result would be an augmented instrument, combining an acoustic instrument and a digital gestural controller (Miranda/Wanderley.2006: 21).
Figure 1: Model representing the augmented trumpet. Different gestures from the performer control the acoustic and electronic parts of the instrument. These two routes are combined in the Sound/effects engine to merge into one output audio signal.
Figure 2: Representation of the Controller.
Figure 3: Representation of Sound Engine.
These representations are based on the model for DMI´s presented by Alexander Refsum Jensenius in the lecture “Musikalsk Elektronikk” (Jensenius.2013)
a. ”Decide on gestures”
The first step suggested by Miranda/Wanderley is to decide what gestures will be used to control the system. The definition on gestures is here set to: “any human action used to generate sound” (Miranda/Wanderley 2006:5). I decided to use the possible movement from fingers that are usually unoccupied when playing the trumpet, and started with these four: left thumb, left index finger, left pinky and right thumb. The instrumental gestures these fingers would execute would be what Calude Cadoz referred to as ergotic gestures, as there would be “energy transfer between the hand and the object”. An exception would be the instrumental gesture performed by the right thumb on the touch sensor. This would be an Epistemic Gesture as it is based on “ our capacity of touch” rather than the previously mentioned “energy transfer”
(Miranda/Wanderley.2006: 9). All of these would further be based on direct gesture aquistion as the sensors directly monitor the actions, or instrumental gestures, of the performer (Miranda/Wanderley.2006: 12).
b. ”Define gesture capture strategies”
In this step one decides how to best translate the gestures from step a. into electrical signals. What kind of sensors will be used and which gesture variable will be captured? In this step I will also discuss the development process of the prototype hardware. In addition to decisions regarding sensor type and gesture capture, this process also included challenges related to sensor placement and stability.
The first picture is an early concept sketch exploring possibilities for sensor placement and interaction possibilities. The second is another early sketch illustrating the potential chaos of signal cables. Although this does not pose a significant challenge with the prototype, one might be faced with this challenge given the addition of more sensors in later versions. A solution could be wireless transfer of signal data to the sound engine through an IPhone application. This could, however, result in latency problems. Minor differences in timing may not critical in more open soundscapes, but for the CAS Trumpet Mod to be functional in extended horn sections, the effects triggers need to be as accurate as absolutely possible. I therefore prefer to keep the sensors and the sound engines connected by physical wires as long as this does not significantly affect performance mobility.
This is a crude sketch showing practical placement of the different sensors. Although there has been made some changes, the prototype is based mostly on this layout. Button 1 and Button 2 in the sketch were initially on/off button switches. In the prototype force sensors replaced these, though still set in the same positions. The switch triggered by the right thumb (top left in sketch) was in the prototype development replaced by an on/of touch sensor as I wished to explore the capabilities of this sensor type. Only the slider mounted near the lead pipe was implemented without alterations.
I borrowed a PhidgetsInterfaceKit from the Institute of Musicology and used sensors from this kit as the hardware basis of the CAS Trumpet Mod Prototype. All sensors to be used for the prototype were contact sensors, as they needed to be in physical contact with the source of energy (read fingers) for the stimuli to be converted into electrical signals. (Miranda/Wanderley.2006: 105). Two force sensors were to be used to register the gestures from the left thumb and the left pinky respectively. A slider was to be used to register the position of the left index finger, while an On/off Touch sensor captured the gesture of the right thumb.
The next task was to check whether the positions planned in the sketches could work, and if so, how to attach the sensors to the trumpet. The sensors needed to be stable in position when attached to allow the best possible functionality of the augmented instrument. But they also needed to be easy to remove and not to damage the instrument in any way, as I wanted to be able to continue to use the trumpet traditionally as well.
Force Sensor no.1
It was important that the force sensors could be manipulated without inhibiting the functionality of the other fingers. The finger responsible for manipulating the sensor also needed to be able to execute adequate pressure on the sensor, so that the mapped effects/sounds could be manipulated expressively. I am here testing the placement of the first force sensor, and how executing force on the sensor affects the movability of the other fingers. I found that the thumb could be able to execute relatively large amounts of pressure on the sensor without affecting the other fingers, as long as the palm rested on the backside of the valve casings. Note also that the left pinky is unoccupied.
To achieve more stability, I decided to mount the sensor on a small plate. Cardboard was used for all plates as it is cheap, easy to work with, does not scratch the trumpet, and hard enough to achieve the desired stability. For the more permanent parts for the CAS Trumpet Mod v.1 I am considering either smooth edged plastic plates or 3D printed parts. The latter method has been successfully used by by Onyx Ashanti to create his “Beatjazz” controllers (Ashanti).
The attachment mechanism was built fastening Velcro and a piece of rubber tape to the cardboard plate using Multi Tac Putty and staples. The Velcros function was to enable the sensor to be attached around the middle valve tube as shown below. The piece of rubber tape provided the necessary friction for the sensor to stay in place while also functioning as a protecting layer between the staples and the valve casing.
The sensor mounted on the attachment mechanism. Although consisting of several layers, the combined part turned out quite slim.
The Velcro attachment on the backside of the second valve tube. Note that the attachment mechanism does not pose a hindrance for the left hand position.
The attachment mechanism for the on/off touch sensor proved to be more of a challenge. The sensor was to be mounted underneath the lead pipe and the final pipe before the bell and I therefore had to build a somewhat more complex cardboard plate as shown above. This resulted in the combined part shown attached in the pictures below. Note the easy access from below to the touch sensor plate.
The output slot on the upper side of the sensor was also part of the reason for the increased complexity its attachment mechanism. The output slot had to be fitted between the pipes while still avoiding the supporting stem.
Force sensor no.2
The Force sensor no.2 was to be controlled by the left pinky,
and the attachment mechanism for this was made nearly identical
to that of Force sensor no.1
Note the small difference in inclination of the sensor to provide increased
accessibility. Both force sensors and the touch sensor shown mounted below.
The sensors still pose no hindrance to the positioning of the left hand.
Checking the position of the slider to be controlled be the left index finger.
The sensor was to capture the finger position and thus it was important to
secure finger mobility.
The slider was attached using the same Velcro/rubber tape system as the
previous sensors. I also attached a metal ring to the slider knob inspired
by that of the third valve tube of the trumpet. This would make it easier for
the sensor to register finger movements.
The cardboard plate added on top provided increased stability as well as a more elegant look from the outside angle.
This was to be the ADC of the system and it needed to be in relative proximity to the sensors to avoid unnecessary cable chaos. I therefore decided to mount it on a cardboard plate, which could then be attached to the right hand of the performer using Velcro. I added two more Velcro bands further down the USB cable so this one could be attached to the right forearm of the performer.
Finished Prototype Hardware
c. ”Define sound synthesis algorithms that will create the sounds to be played”
In our case this step could also be rephrased as “Define the sound effect algorithms that will manipulate the input audio”. In principle, the signals from the sensor system described above could be mapped to any sound or effects engine. For the Prototype, I decided to focus on input audio manipulation, or sound effects. This way the acoustic and electronic elements leading up to the final output sound would be both clearly separable, and at the same time perceived as one enhanced instrument. This could make it easy to analyze how the interaction with the electronic elements affects the total performance, and evaluate the interactive characteristics of the augmented instrument.
This step includes programming the sound engine for the DMI. The sound engine was developed in Max (MSP) using a number of different patches. I used a MacBook as digital platform.
For MAX to be able to interact with the PhidgetsInterfaceKit, it was first necessary download the externals for the PhidgetsInterfaceKit and to introduce these into Programs à MAX6 à Cycling ´74 à max –externals.
Sound Engine 1
This sound engine was programmed to provide a basic setup for testing and exploring the functionality of the prototype controller. The effects used were an overdrive and a stereo delay. These are effects that I have considerable experience with, thus making the interaction with the controller the main focus of the setup. I used the following approach programming:
cmd+ n : New document
cmd+ e : Edit mode
M “Start” Connected to inlet 1 of “PhidgetInterfaceKit”
M “Stop” Connected to inlet 1 of “PhidgetInterfaceKit”
M “getVersion” Connected to inlet 1 of “PhidgetInterfaceKit”
M “getStatus” Connected to inlet 1 of “PhidgetInterfaceKit”
M “getSerial” Connected to inlet 1 of “PhidgetInterfaceKit”
M “read” Connected to inlet 1 of “PhidgetInterfaceKit”
M “setSamplerate $1” Connected to inlet 1 of “PhidgetInterfaceKit”
I Connected to message of “setSamplerate $1”.
N “route di ai” Connected to output from “PhidgetInterfaceKit”
N “unpack 0 0 0 0 0 0 0 0” Connected output from “di”
I connect to “unpack 0 0 0 0 0 0 0 0” . For each of the 4 elemtents used.
Element 8: Slider, Element 7: Force Sensor no.1, Element 6: Force Sensor no.2, Element 5: Touch Sensor
Sound system module:
Analog to Digital Converter Collecting Sound
M “start” connected to inlet of “adc~”
M “stop” connected to inlet of “adc~”
M “startwindow” à connected to inlet of “adc~”
M “open” connected to inlet of “adc~”
M “wclose” connected to inlet of “adc~”
T connected to inlet of “adc~”
N “meter~” connected to “Audio In Ch 1” of “adc~”
N “overdrive~” connected to “Audio In Ch 1” of “adc~”
N “overdrive~” connected to “Audio In Ch 1” of “adc~”
N “*~1” connected to outlet of “overdrive”, ”overdrive” and to connected to “Audio In Ch 1” of “adc~”
The overdrive was set directly after the ADC and before the delay to enhance signal clarity.
N “delay~ 45100” connected to outlet of “*~1”
Nà “ *~ ” connected to outlet of “delay~ 45100”
N “delay~ 46100” connected to outlet of “*~1”
N “ *~ ” connected to outlet of “delay~ 46100”
N “dac~” left inlet connected to outlet of (“delay~ 45100” through “*~”) and outlet of “*~6”
right inlet connected to outlet of (“delay~ 46100” through “*~”) and outlet of “*~6”
The “adc~” collected audio from the microphone plugged into input 1 in M-Audio Fastrack Pro external sound module. This was achieve by changing the settings in Max through Options à Audio Status à Input/Output device. For connection between PhidgetsInterfaceKit module and Sound System module see step.d “Mapping 1” below.
d. ”Map the sensor outputs to the synthesis and music control inputs”.
Mapping is how the variables from the sensors relate to the parameters of the sound/effect engine (Miranda/Wanderley.2006: 14). This process includes filtering, scaling and segmenting the signals (Jensenius.2013).
To keep the interaction clarity needed to evaluate the efficiency of the hardware/controller, I chose to mostly use the mapping strategy one-to-one. This means mapping each one gestural variable to one effects processor parameter (Miranda/Wanderley.2006: 16). All the sensors from the PhidgetsInterfaceKit had a range of 0-999. These parameters therefore had to be scaled in order to fit the scales of the different effect parameters.
The parameter from the slider was mapped to control the delay time of the stereo delay. The delay time is set in samples by default, and maximum delay time was set to 45100 and 46100 samples. To achieve the stereo effect, I scaled the signal going to the left channel by a factor of 50 and the signal going to the right channel by a factor of 100. Not only would the delay time at a given gesture parameter signal between the left and right channel differ, but the difference would also vary in linear relationship to the increased/decreased gesture parameter signal. Even though this mapping initially can be seen as one-to-one, it may also be characterized as one-to-many (Miranda/Wanderley.2006: 16) if you consider right delay time and left delay time two different effect parameters.
The parameter from Force Sensor no.1, controlled by the left thumb, was mapped to control the output volume of the stereo delay. The output was downscaled by dividing the signal by 100 (Nà “/ 100”), thereby getting “*~” values between 0 and 9.99. The goal of this mapping was to achieve an expressive output.
The parameter from Force Sensor no.2, controlled by the left pinky, was also downscaled by a divisor of 100 (Nà “/ 100”). It was then mapped to control the volume one of the overdrives.
Upon experimenting with the output of the touch sensor, I found it difficult achieve satisfying results linking the output signal directly to one effect parameter. The sensor constantly sends an output of 999, which drops to 0 upon touch. Using inverse scaling, this could be turned around, making the sensor output trigger a constant “on” state of a parameter. Instead, I decided to use the sensors inherent property for signals dramatically drop. I scaled the signal adding 2 (Nà “+ 2”) and dividing my 200 (Nà “/ 200”). As I used Integer number boxes this mean that the output number would be an integer between 0 and 5. Upon touch the sensor would then send a signal causing one of the active overdrive effects to drop silent, resulting in a “sudden silence” effect.
e.”Decide on the feedback modalities available”
These feedback modalities can be visual, tactile and/or kinesthetic (Miranda/Wanderley.2006:4).
The electronic sensors produce only barely audible passive or primary feedback (Miranda/Wanderley.2006: 11), and thus most of the feedback within these categories derives from the noises produced in normal use of the trumpet. Be it the noise from triggering a valve or the click of the third valve tube, this feedback is of little practical importance, as these noises are easily drowned out by the amplification of the system. Other primary feedback from the instrument includes the kinesthetic feedback when executing different gestures on the sensors. Here, the sensors do not actively respond, but the feedback is linked to the passive qualities of the sensor materials. (Miranda/Wanderley.2006: 71) The touch sensor feels smooth, the pressure sensors feel hard and a bit sharp when pressed, and the metal ring of the slider is feels solid and a bit cold.
Both the visual feedback and the passive feedback could convey a sensation of being interconnected with the computer. This sensation is could further be enhanced through the secondary, or active feedback, since the physical actions of the performer manipulates both digitally processed and acoustic sounds.
System Test and Evaluation
The CAS Trumpet Mod Prototype was demoed the first time live at the Institute of Musicology the 30 of October 2013. All parts of the system were functional at this time, but adjustments have later been made based on the following observations.
At the concert, Hilde Marie Holsen and I performed a live improvised electro-acoustic soundscape. She played trumpet with Ableton live while I played the CAS Trumpet Mod Prototype. The one-to-one based mapping of Mapping 1 made the control of the CAS Trumpet Mod almost intuitive, and I found that the effects of Sound Engine 1 functioned well in the particular musical setting.
Of the effect parameters, I found the delay volume controlled by Force sensor no.1 to be the most expressive. In combination with longer tones of the trumpet, this effect resulted in the sensation of being able to explode and diminish the auditory landscape upon sensor stimulation. This was also a natural result of the sensitivity of Force Sensor no.1, as smaller variations in delay volume were more difficult to control.
The expressivity of the slider parameter was opposite in comparison. The slider itself was relatively slow and I also found that the index finger movement was restricted to a few centimeters. Delay time controlled by this sensor/gesture combination thus worked best providing nuances in the expressions of Force Sensor no.1. The exception was when more impulsive sounds (Nymoen 2013) were played, such as trumpet stabs. Here the variations in delay time provided communicative musical effects.
Force Sensor no.2 was found quite difficult to operate in combination with other instrumental parameters. This was an effect of the previous scaling of the mapped sensor signal, as considerable pressure was needed in order to get efficient musical results from this overdrive volume parameter. As excessive force by the pinky compromises the movability of the other fingers, this musical parameter was most efficiently used in combination with no other manipulation of the electronic system elements. I therefore found it necessary to upscale the signal to the current scaling after the concert to provide a more balanced usability of the sensor parameter.
The “sudden silence” effect provided by the touch sensor proved the least intuitive of the available musical elements. The use of this element was therefore quite limited during the first demonstration/concert. Later tests of the system however, have found the effect to be quite expressive given the right timing.
With the exceptions discovered at the first trial, further tests of the system have found the overall sensor placement to be quite efficient, and the controller/sound engine mapping to be clear and expressive. To execute efficient and nuanced control of the different electro-acoustic parameters simultaneously takes practice, even though the initial interaction with the prototype setup is quite intuitive. Adding the electronics variables causes the overall trumpet playing to be more focused on timbre and simple phrases/sounds in combination with the effects, creating a combined expressive audio output. The clarity of the first effect/sound engine and mapping combination makes it useful for making a clear and simple statement in an ensemble. To provide more nuances, complex sound engines and mapping strategies will be used in the future, but Sound Engine 1 and Mapping 1 may still prove the most efficient in more traditional band settings.
Upon researching augmented instruments as part of the sketching process, I became aware of the “meta-trumpet” presented by Jonathan Impett at the ICMC ´94 proceedings (Impett). The lecturer of MUS2830, Kristian Nymoen, also told me about the “electrumpet” developed by Hans Leeuw (Leeuw). Although the basic outline of the CAS Trumpet Mod Prototype was developed by the time I read up on these augmented trumpets, the “meta-trumpet” and the “electrumpet” have provided me with inspiration and ideas for the further development of the CAS Trumpet Mod. Following are a few notes regarding the CAS Trumpet Mod Prototype in relation to these two.
To start with the obvious, all three are augmented instruments where sensors have been added to a trumpet to capture additional gestures. Both the “meta-Trumpet” and the CAS Trumpet Mod uses two pressure sensors placed on the valve casing. On the “meta-trumpet” both are placed to the right of the third valve casing, while on the CAS Trumpet Mod one is place there while the other is placed on the right of the first valve casing. The “meta-trumpet” additionally uses ultrasound transmitters, mercury switches, magnetic field sensors and regular switches thus being a further developed augmented instrument.
Both the “electrumpet” and the CAS Trumpet Mod is designed with focus on not compromising the normal playing position of the fingers. For the former, this is stated in the documentation for website of Hans Leeuw (Leeuw). Additional sensors have also been added on this instrument, including a second mouthpiece for air-pressure control, slider buttons, pressure sensors and switches.
An important difference between the CAS Trumpet Mod Prototype and the two other augmented trumpets presented here is found in sensor detachability. When designing the CAS Trumpet Mod Prototype, the ability to detach the sensor system from the trumpet has been an important focus to allow the performer a continued choice between traditional and augmented instrument. From the documentation referred below, it does not appear that this has been a focus in the development of either the “electrumpet” or the “meta-trumpet”.
The “meta-trumpet” and the “electrumpet” have served as a guide for what is possible to achieve through further development of the CAS Trumpet Mod. Both of the instruments are currently more developed than the CAS Trumpet Mod, with complex sensor systems and physical modifications. This gives the instruments several additional “dimensions of sound” to operate within, as many more sound/effects engine parameters can be controlled by the performer.
These abilities will serve as further inspiration in the continued development process. I do however consider it a definite advantage being able to detach the system completely from the regular trumpet. This allows the performer at any time to chose whether to be a “meta-trumpeter” or not. As long as the attachment mechanisms are flexible, this also allows the Mod to be used with different trumpets. I therefore doubt I will make changes to the CAS Trumpet Mod that will compromise this ability.
Thoughts on further development
As the PhidgetsInterfaceKit is borrowed from the University, I will need to acquire other sensors for use in the CAS Trumpet Mod v.1. Based on the evaluation above, I will most likely keep the positioning and type of the two force sensors as well as the slider, though preferably with somewhat smaller components. Whether the touch sensor will be kept on in the v.1 is uncertain, as the mapping possibilities in fitting several on/off switches in its place is intriguing.
The CAS Trumpet Mod v.1 is going to a dynamic system. First, means keeping the dynamic, interactive elements explored in the prototype. Secondly, it means building a flexible hardware system allowing for changes in sensor type and placement along with the varying needs of the different musical settings.
More permanent material solutions are going to be used, such as the mentioned 3D-printing option. I would also like to experiment with biometric sensors by mapping the pulse of the performer to different sound engine parameters.
As the first sound engine explored basic sound effects the next ones will explore new effect combinations and sound synthesis options. In the prototype, the sound engine functioned as an extension of the trumpet sound. In the v.1 sound engines, I wish to combine this “extension approach” with sounds completely separate from the trumpet, thereby creating an interaction between the pure electronic elements, the electro-acoustic elements and the purely acoustic elements of the instrument. These types of sound can be used to transform the performer from an “enhanced instrumentalist” into more of a “soundscaper”.
I also wish to explore more complex mapping strategies, resulting in soundscapes that are seemingly self-sustained. This can be achieved by utilizing many-to-one mapping (Miranda/Wanderley. 2006: 16) and by implementing a more indirect relationship between the sensor gesture and the sound engine parameter. An example of this could be the direct signal from a sensor controlling one parameter of a synthesis process, while the differentiated signal could control another. Mastering such a mapping technique will require all the more practice, but the end result can be both highly expressive and precisely nuanced.
Through this assignment, I have successfully developed a functioning prototype for the CAS Trumpet Mod and explored several different challenges related to the musical interaction with this controller. The experiences and skills acquired in this process will be applied in the development of the CAS Trumpet Mod v.1.
Ashanti, Onyx. Onyx Ashanti Webpage.
http://onyx-ashanti.com/. Downloaded 8.12.13.
Impett, J. 1994. A meta trumpet(er). Proceedings of the 1994 International Computer Music Conference (ICMC ´94) .Aarhus, Danmark, pp. 147-50. San Francisco: ICMA.
Jensenius, Alexander R. 2013. “Musikalsk Elekronikk”. Lecture in MUS2830- Interactive Music. Fall. University of Oslo, Oslo
Leeuw, Hans. Electrumpet Website.
http://electrumpet.nl/Site/Electrumpet.html Downloaded 8.12.13.
Miranda. E.R ,Wanderley, M.M. 2006. New Digital Musical Instruments: Control and Interaction Beyong the Keyboard. A.R Editions Inc. ,Middleton, Wisconsin.
Nymoen, Kristian. 2013. “Oppsummering MUS2830 Høst 2013”. Lecture in MUS2830- Interactive Music. Fall. University of Oslo, Oslo
Svalesen, Christian Aa. 2013. Chasm Music Website
All digital software has been downloaded from or through the MUS2830- Interactive Music page at the University of Oslo website.