Theremin-Playing Robot

The Theremin

The Theremin is an electronic musical instrument which is played without physical contact. The Theremin was invented in 1919 by Lev Sergeyevich Termen, who later changed his name to Leon Theremin. A typical Theremin resembles a box with two protruding metal antennae, one on either side. A single musical tone is controlled through interactions with the antennae. One antenna controls pitch, the other volume. As the musician's hand approaches the pitch antenna, the pitch of the sound increases (Fig.1 (a)). As the other hand approaches the volume antenna, the volume of the sound decreases(Fig.1 (b)). Most Theremins are capable of producing a 5 octave range.

Figure 1: Controlling the Sound from the Theremin

The Theremin is difficult to play for severalreasons. A Theremin is much like a trombone in thatthere is no exact "finger position" to give the desired note. Themusician must deduce the correct pitch within the context of the notespreviously played. Since the player never touches the instrument, there is alsono physical point of reference to associate with specific tones. In addition,the response of the capacitive antennae can vary with temperature and humidity.Only a handful of people have ever mastered the Theremin.

The Theremin-Playing Robot

Fig.2 shows the system overview. A human plays notes on the keyboard, which sends information to the Theremin-Playing Module (TPM) using the General MIDI format. This information indicates which note has been played. The TPM uses this code to determine the desired pitch, or frequency, in Hertz. The keyboard also sends messages indicating the volume of the note, as well as control messages that affect the motion of the robot's arms. The TPM samples the audio signal from the Theremin, and using the desired pitch and volume information, computes position commands to send to the Arm Controller modules. These modules move the arms? end-effectors to the desired positions.

The core of the TPM is an audio-servo loop (Fig.3). This loop moves the robot's right arm (the Pitch arm) such that the Theremin emits a note of a desired pitch. The audio signal from the Theremin is sampled by the PCs sound card at 8kHz. The sampled audio data is then transformed to frequency information using a Fast Fourier Transform (FFT). 512 data-points, zero padded to 2048, are transformed each time. This provides a spectrum with frequency resolution of 3.90625 Hz. The frequency component with the largest magnitude is then taken to be the fundamental frequency of the Theremin's audio signal, represented as f in Fig.3.

Figure 2: Theremin-Playing Robot System

Figure 3: Audio Feedback Loop

In addition to piano keys, the MIDI keyboard used in the system also has several control knobs. These also generate MIDI messages which the TPM uses to vary certain aspects of the robot's play. One of these control knobs controls the dynamics of the audio servo loop. Two other control knobs are used to add vibrato to the music. In this mode, a sinusoidal signal is added to the robot's joint control signals. The control knobs vary the amplitude and frequency of the sinusoidal signal.


As mentioned previously, the Theremin is a difficult instrument to play. The Theremin's sensitivity to environmental changes requires the player to have "perfect pitch"---that is, the ability to identify exactly the note being played by an instrument. The TPM effectively gives ISAC ``perfect pitch.'' The performance of the robot then becomes function of its servo control bandwidth. ISAC's softarms are not ideal in this respect; their control bandwidth is quite limited. However, the arms are human-like in appearance and therefore may be more appealing to watch from an entertainment perspective. Additionally, the flexing of the muscles during the oscillatory vibrato is quite natural appearing and easy to obtain with a rubbertuator-based humanoid robot. Electro-mechanical arms can vibrate (although it is not conducive to longevity) but they are very stiff and ``robot-like'' in high frequency movement. A video of ISAC playing the Theremin was shown at a Theremin Festival in Portland, Maine, in 1997. The conference attendees were mainly Theremin musicians with an appreciation for the difficulties in playing the Theremin. They enjoyed the robot's perfect pitch and were amused by the vibrato---vibrato is often over-used by beginning Thereminists to cover up ``sour'' notes.


A. Alford, S. Northrup, K. Kawamura, K-W. Chan, and J. Barile. "Music Playing Robot." Proceedings of the 1999 International Conference on Field and Service Robotics (FSR '99). August 29-31, 1999. Pittsburgh, PA, pp. 174-178.