Voder: Difference between revisions

From MSc Voice Technology
Jump to navigation Jump to search
Line 10: Line 10:
Before Voder, several significant milestones have marked progress in recreating human speech artificially.  
Before Voder, several significant milestones have marked progress in recreating human speech artificially.  


 
The earliest attempts towards speech synthesis using mechanical means can be traced back to 1779 when [[Mechanical synthesis#Christian Kratzenstein, 1779: generating vowels with acoustic resonators|Christian Gottlieb Kratzenstein]] [http://research.spa.aalto.fi/publications/theses/lemmetty_mst/image48.gif<nowiki>] </nowiki>produced the five vowel sounds /a/, /e/, /i/, /o/, /u/ using various shaped tubes<ref>Oggerino, J. J. (1980). An evaluation of a talking machine, the HC 120 Phonic Mirror Handivoice.</ref>. A few years later, W. R. von Kempelen of Vienna further advanced this and managed to produce not only vowel sounds but also a number of consonant sounds. His [[Wolfgang von Kempelen's Speaking Machine (1769)|Speaking Machine]], dating back to the late 18th century, utilized bellows and reeds to simulate limited vowel and consonant sounds, underscoring the potential for creating full-fledged artificial speech.  
The earliest attempts towards speech synthesis using mechanical means can be traced back to 1779 when [[Mechanical synthesis#Christian Kratzenstein, 1779: generating vowels with acoustic resonators|Christian Gottlieb Kratzenstein]] [http://research.spa.aalto.fi/publications/theses/lemmetty_mst/image48.gif<nowiki>] </nowiki>produced the five vowel sounds /a/, /e/, /i/, /o/, /u/ using various shaped tubes<ref>Oggerino, J. J. (1980). An evaluation of a talking machine, the HC 120 Phonic Mirror Handivoice.</ref>. A few years later, W. R. von Kempelen of Vienna further advanced this and managed to produce not only vowel sounds but also a number of consonant sounds. His [[Wolfgang von Kempelen's Speaking Machine (1769)|Speaking Machine]], dating back to the late 18th century, utilized bellows and reeds to simulate limited vowel and consonant sounds, underscoring the potential for creating full-fledged artificial speech.


Charles Wheatstone built upon Kempelen's work with his Speaking Machine in the early 19th century. By incorporating a vibrating reed, [[Wheatstone's "Speaking Machine" (1824)|Wheatstone's machine]] [http://research.spa.aalto.fi/publications/theses/lemmetty_mst/image49.gif<nowiki>] could produce a wider range of sounds compared to previous attempts, resulting in more accurate and recognizable speech sounds</nowiki><ref>Dudley, H., & Tarnoczy, T. H. (1950). The speaking machine of Wolfgang von Kempelen. ''The Journal of the Acoustical Society of America'', ''22''(2), 151-166.</ref>.
Charles Wheatstone built upon Kempelen's work with his Speaking Machine in the early 19th century. By incorporating a vibrating reed, [[Wheatstone's "Speaking Machine" (1824)|Wheatstone's machine]] [http://research.spa.aalto.fi/publications/theses/lemmetty_mst/image49.gif<nowiki>] could produce a wider range of sounds compared to previous attempts, resulting in more accurate and recognizable speech sounds</nowiki><ref>Dudley, H., & Tarnoczy, T. H. (1950). The speaking machine of Wolfgang von Kempelen. ''The Journal of the Acoustical Society of America'', ''22''(2), 151-166.</ref>.

Revision as of 12:44, 18 October 2023

Introduction

The Voder, short for "Voice Operation DEmonstratoR," stands as a pioneering achievement in the realm of speech synthesis technology. Conceived and developed by an American electronic and acoustic engineer Homer Dudley in the 1930s at Bell Labs, the Voder was a groundbreaking invention that laid the conceptual groundwork for subsequent advancements in the speech synthesis field and paved the way for the sophisticated Text To Speech technologies we rely upon today.

Designed similarly to an organ – a musical instrument with pedals and keys – Voder demanded an operator to manipulate its components effectively: through this manipulation, the machine could approximate speech-like sounds, portraying a semblance of human communication.

Desite the fact that the voice quality was rather limited and the machine required months of training to master, the invention of Voder has sparked interest and paved the way for speech synthesis. In fact, the structure of Voder and the application of filter-source model heavily influenced the later developments.

Historical Context

Before Voder, several significant milestones have marked progress in recreating human speech artificially.

The earliest attempts towards speech synthesis using mechanical means can be traced back to 1779 when Christian Gottlieb Kratzenstein [http://research.spa.aalto.fi/publications/theses/lemmetty_mst/image48.gif] produced the five vowel sounds /a/, /e/, /i/, /o/, /u/ using various shaped tubes[1]. A few years later, W. R. von Kempelen of Vienna further advanced this and managed to produce not only vowel sounds but also a number of consonant sounds. His Speaking Machine, dating back to the late 18th century, utilized bellows and reeds to simulate limited vowel and consonant sounds, underscoring the potential for creating full-fledged artificial speech.

Charles Wheatstone built upon Kempelen's work with his Speaking Machine in the early 19th century. By incorporating a vibrating reed, Wheatstone's machine [http://research.spa.aalto.fi/publications/theses/lemmetty_mst/image49.gif] could produce a wider range of sounds compared to previous attempts, resulting in more accurate and recognizable speech sounds[2].

In the late 1920s, the author of Voder had created the much more well-known Vocoder[3], which coded human speech across telephone lines by turning incoming speech into electronic signals and then replicated it on the other end using electric sounds meant to mimic a person’s voice. The Voder, in turn, went one step further: it produced speech without the input of the human voice.

Working Mechanism

VODER was operated by a skilled human operator who used a keyboard, switches, and a foot pedal to produce speech sounds.[https://upload.wikimedia.org/wikipedia/commons/c/ca/Homer_Dudley_%28October_1940%29._%22The_Carrier_Nature_of_Speech%22._Bell_System_Technical_Journal%2C_XIX%284%29%3B495-515._--_Fig.8_Schematic_circuit_of_the_voder.jpg] The operator had to internalize a set of rules (typically a year or more of training was required)   in order to produce intelligible speech, and each utterance had to be carefully rehearsed.

The operator can manipulate the resonances of the vocal tract, thereby affecting the sound characteristics. Factors like voicing and pitch can also be controlled by levers and pedals. The process of making the VODER speak can be almost perceived as playing an instrument.

In VODER, There are 10 filter circuits in the motor and combined with the two energy sources they give a total of 20 separate components to be used in building up speech sounds. A set of manual controls was also supplied. The output amplitudes of the filters were controlled from a keyboard. A wrist bar selected buzz or hiss excitation (as shown in the graph). A foot pedal  controlled fundamental frequency.

The VODER consisted of  a wrist bar for selecting a voicing or noise source and a foot pedal to control the fundamental frequency. The source signal was routed through ten band pass filters and their output levels were controlled by fingers.[http://research.spa.aalto.fi/publications/theses/lemmetty_mst/image50.gif]

The Voder is a manually operated speech synthesizer that recreates the physiological characteristics of the human voice. It worked by breaking up human speech into its acoustic components using a set of ten contiguous band-pass filters that covered the entire speech frequency range and were chosen after a careful analysis of how the human ear interprets speech sounds. The initial sounds produced by either the oscillator or the gas discharge tube were passed through these filters, and their outputs were then passed through an amplifier that mixed and modulated them and passed it on to a loudspeaker in order to produce an electronic human speech[4].

Thus, The Voder was designed on the principle that an electrical speech synthesizer must be a functional copy of the human speech source-filter mechanism that consists of the following steps:

  1. Initiation: having a steady energy source. This is achieved by the compressed air in the lungs. Correspondingly, the Voder had an electric supply for its steady energy source.
  2. Phonation: converting the energy into audible vibrations. This is achieved by two distinct methods: voiced sounds are produced by the vibration of the vocal cords in the larynx, and unvoiced sounds and whispers are produced by forcing a stream of air through a constriction in the vocal tract, which produced a hissing sound. Likewise, the electrical oscillator in the Voder produced the buzz tone that was used to create vowels and nasal sounds, and the ‘random noise’ source made the hissing sound used for producing fricatives.
  3. Articulation: creating various speech sounds. This is achieved by adjusting the resonances and couplings of air cavities through the movements of organs such as jaws, lips, and tongue. The air in these cavities resonates in a specific fashion to make certain overtones produced by the larynx to be radiated more efficiently to the atmosphere than others. In the Voder, this was achieved with the help of electrical filters. The potentiometers (devices that control how much electricity flows through a circuit) controlled by the finger keys were used to operate these filters.

Pitch control, in turn, was managed by a foot pedal, which also converted tones and hissing sounds into vowels, consonants, and inflections. Three additional keys were dedicated to triggering transient excitation of selected filters to produce plosive sounds, such as 'p' or 'd,' as well as affricative sounds, like 'j' or 'ch.'

Overall, the Voder machine was designed with ergonomics in mind, intended to function within the capabilities of a human operator. A normal human operator would only be able to control a maximum of ten keys at a time, and hence the spectrum keys were limited to ten. In practice, Voder was not an easy machine to work with, and it could take more than a year to become a professional operator of the machine[5].

Key Innovations

The Voder was among the first devices to allow manual control of speech synthesis. It was a pioneer in electronic sound generation, breaking down human speech into its fundamental acoustic components and reproducing these patterns electronically. This was a significant advancement in the early stages of electronic speech synthesis. Moreover, Voder was the first successful attempt at recreating an important physiological characteristic of the human voice – the ability to create voiced and unvoiced sounds[6].

To improve the operator's performance, the Voder had a recording and playback feature that allowed operators to objectively analyze their areas of improvement. This feature is similar to modern-day contact centers that use call recording and analysis to improve agent performance.

In fact, the abilities of Voder went beyond human voice, as it could also produce non-speech sounds such as musical tones and sound effects[7], and thus it was used in a variety of applications, including radio broadcasts, sound effects for movies, and even music performances.

Impact

The Voder was demonstrated to the public at the 1939 New York World's Fair, attracting widespread attention and showcasing the possibilities of artificial speech production. It was a significant step towards public awareness and interest in the field of speech synthesis. Bell Labs also continued its research in this field, and nearly two decades later, in the 1960s, they created a robot that could sing[8].


Despites the speech quality and intelligibility of VODER were far from good, the capability for producing artificial speech was well demonstrated, as VODER shows how intelligible speech can be produced artificially. In fact, the basic structure of VODER is pretty similar to systems based on the source-filter-model of speech in the modern days.

Other electric models back in the 1950s and 1960s used a filter to pass the electric source along and emulated the resonance of the vocal tract as well. The articulation can be categorized into two methods: transmission line analog and terminal analog. The former one uses numerous circuits that represent sections of the vocal tract, while the later one simulates the vocals regardless of the shape of the circuits.

Ever since the 1970s, the automatic text-to-speech systems have commercialized speech synthesis. The first system was created by the team led by Noriko Umeda in Japan back in 1968.[http://research.spa.aalto.fi/publications/theses/lemmetty_mst/image51.gif]

Future Research

For future research, it would be beneficial to provide a more comprehensive discussion of the challenges and limitations faced by the Voder. Exploring the difficulties operators encountered, the complexities of mastering this machine, and the constraints in terms of the range and naturalness of speech it could produce would offer a deeper understanding of the technology's historical context. This could include the descriptions of experiences from those who operated the Voder with explaining the training process that operators underwent to become proficient in manipulating the keys, filters, and other components effectively.


There are two competing paradigms for speech synthesis, which can be broken down into articulatory synthesis and terminal-analog synthesis. Articulatory synthesis focuses on modeling the articulators of the speech production mechanism in detail. While terminal-analog synthesis aims at modeling the speech signal itself, through a source-filter model.

VODER is an example of terminal-analog synthesizers as VODER simulates the acoustic output of the vocal tract (acoustic properties of the speech) but not the activity of the vocal tract itself.

Modeling the signal (terminal-analog synthesis) is intrinsically simpler than modeling the articulators (articulatory synthesis). The terminal-analog synthesizers were prevalent, and articulatory synthesis still remains an active field of research.

LLM Review

We asked ChatGPT to act as an experienced professor in the field of voice technology to review our wiki page assignment.This was the prompt we used: "Imagine you're an experienced editor and distinguished professor in voice technology, tasked with refining our group's Wiki page on speech synthesis. Your aim is to enhance the formality of the content while ensuring clarity and engagement for the readers. Please review the Wiki page we've created and provide feedback on any errors, areas for improvement, or suggestions for enhancing the overall quality of the article."

Introduction: The introduction is well-written, providing a clear overview of the Voder and its significance in the realm of speech synthesis technology. It effectively sets the context and lays the foundation for the subsequent sections.

Historical Context: The historical context provides a good overview of key milestones in speech synthesis technology. It effectively traces the development of speech synthesis from earlier attempts to the creation of the Voder and its predecessor, the Vocoder

Working Mechanism: The section explains the working mechanism of the Voder in a detailed and structured manner. It effectively breaks down the steps involved in producing speech-like sounds using the Voder and compares it to the physiological characteristics of the human voice.

Key Innovations: This section effectively highlights the key innovations of the Voder, emphasizing its manual control of speech synthesis, electronic sound generation, and the ability to create voiced and unvoiced sounds. The mention of the recording and playback feature adds depth.

Impact: The impact section succinctly discusses the significance of the Voder and its public demonstration at the 1939 New York World's Fair. It accurately portrays the attention and interest it garnered, showcasing the potential of artificial speech production.

Future Research: This section wisely suggests future research areas, providing insights into potential directions for a more comprehensive discussion on the challenges and limitations of the Voder. It suggests including experiences from operators and discussing the training process.

Overall Feedback: The content is informative, well-structured, and engaging, effectively covering the history, mechanics, impact, and future research possibilities related to the Voder. The language is clear and concise, making it accessible to readers without compromising on technical accuracy. The chronological approach in the "Historical Context" section helps in understanding the evolution of speech synthesis technology.


Moreover, ChatGPT(3.5) recommended the inclusion of a concluding section to recapitulate the main points and underscore the significance of the Voder in the history of speech synthesis. However, considering that we have addressed this aspect in the 'Impact' section, we did not fully follow the suggestion.

References

  1. Oggerino, J. J. (1980). An evaluation of a talking machine, the HC 120 Phonic Mirror Handivoice.
  2. Dudley, H., & Tarnoczy, T. H. (1950). The speaking machine of Wolfgang von Kempelen. The Journal of the Acoustical Society of America, 22(2), 151-166.
  3. Dudley, H. (1940). The vocoder—Electrical re-creation of speech. Journal of the Society of Motion Picture Engineers, 34(3), 272-278.
  4. Dudley, H. (1940). The carrier nature of speech. Bell System Technical Journal, 19(4), 495-515.
  5. Mills, M. (2012). Media and prosthesis: The vocoder, the artificial larynx, and the history of signal processing. Qui Parle: Critical Humanities and Social Sciences, 21(1), 107-149.
  6. Dudley, H., Riesz, R. R., & Watkins, S. S. (1939). A synthetic speaker. Journal of the Franklin Institute, 227(6), 739-764.
  7. Diesterhöft, S. (2003). Meyer-Eppler und der Vocoder. Seminars Klanganalyse und -synthese, Fachgebiet Kommunikationswissenschaft, Institut für Sprache und Kommunikation, Berlin Institute of Technology.
  8. Cox, C. (1967). The Alien Voice. Alvin Lucier’s North American Time Capsule, 170-186.

Team Members

  • Igor Marchenko
  • Wangyiyao Zhou
  • Yaling Deng
  • Dongwen Zhu
  • Sherry Yu-Ting Yeh