Handbook of Ultrasonic Vocalization: A Window into the Emotional Brain

By Academic Press

Handbook of Ultrasonic Vocalization: Window into the Mammalian Brain, Volume 25, is an exhaustive resource on ultrasonic vocalizations in vertebrates, providing full coverage of all aspects of these vocalizations. The book also demonstrates the usefulness of ultrasonic vocalizations in studies of animal communication, sociobiological states, and in mammalian models of affective disorders, addictions and neurodevelopmental disorders, making it an indispensable resource for researchers using animal models. The book begins with the evolution of vocal communication before discussing mechanisms of ultrasound production, perception and the brain systems involved in emotional arousal that are responsible for the generation of vocalization and emotional states.

In addition, the book covers studies of neuroactive agents and sociopsychological conditions that can regulate the outcome of ultrasonic vocalization and provide clues about animals’ internal states. Critically, the book also includes thorough coverage of pharmacological investigations using ultrasonic vocalizations, increasingly being utilized for studies in affective disorders, psychoses, addiction and alcoholism. No other book provides such extensive coverage of this rapidly growing field of study.

• Represents a multidisciplinary approach that incorporates evolution, communication, behavioral homeostasis, emotional expression and neuropsychiatric dysfunction
• Provides a systematic review of ultrasonic vocalizations in major groups of rodents widely used in laboratory research
• Discusses numerous other species across vertebrates that emit ultrasounds

• Language: English
• Publisher: Academic Press
• Release date: Apr 27, 2018
• ISBN: 9780128097731

Book preview

field.

Chapter 1

Introduction and Overview of the Handbook of Ultrasonic Vocalization

Stefan M. Brudzynski¹; Philip S. Zeskind²,³    ¹ Department of Psychology, Brock University, St. Catharines, ON, Canada

² Department of Pediatrics, University of North Carolina, Chapel Hill, NC, United States

³ Department of Pediatrics, Carolinas Medical Center, Charlotte, NC, United States

Abstract

This introductory chapter describes the importance of the study of ultrasonic vocalizations (USVs), particularly in rodents. Analyses of the emissions of these calls inform us about development, physiological function, and many aspects of social behavior, including emotions and motivation. They also reveal the effects of stress and environmental change on animals. The study of USVs may also be used in the modeling of neuropsychiatric disorders and diseases, the effects of bioactive agents on the brain, the addiction process, and neurodevelopmental deficiencies.

Keywords

Significance of vocalization studies; Handbook overview

I Introduction: Why Study Ultrasonic Vocalizations?

Vocal communication is comprised of a particularly important and unique set of behaviors across animal species. From an evolutionary perspective, the emission of this distal form of contact among conspecifics and others contributes to the individual's survival and development. The basic brain regulation of emission of vocalizations is highly conserved in evolution and could be found in the caudal hindbrain from early fish, trough amphibians, birds, and mammals (Bass, Gilland, & Baker, 2008). Vocal communication may have reached particular importance in early mammals where it has been postulated to have originated in mother-infant interactions along with nursing and play behavior (MacLean, 1985). In later evolution, vocal communication with the mother was modified and maintained through adulthood for exchange with social groups (Brudzynski, 2014). Rodents, which represent the most abundant group of mammals, switched from sonic to ultrasonic vocalizations (USVs) for defensive reasons at later stages of evolution. Through these evolutionary changes, the basic neural regulation of USV emission remained comparable to that of audible-to-human calls and continued to play a strategic and important behavioral role.

Interest in the study of the emission of USVs in rodents in general and laboratory rats and mice in particular has rapidly increased. This interest has its basis in our increased understanding of the critical role USVs play in the social behavior and communication of rodents as well as the realization that vocalizations could be used to study emotional behavior and motivation (Brudzynski, 2014). Further, the elaborate process of emission and the complexity of emitted signals that carry semiotic value (meaning in human terms) have also been a fertile ground for the study of this vocalization. The ability to record and analyze USVs has served well for laboratory studies of the significance of this behavior.

The outcome of this increased interest has now resulted in the study of USV emissions in an astonishing number of behavioral situations, including mother-infant interaction, vocal learning, play behavior, sexual behavior, alimentary behavior, social agonistic or appetitive situations, territorial behavior, alarm, defense, and aggression (Barfield & Thomas, 1986; Brudzynski, 2009; Kaltwasser, 1990; Litvin, Blanchard, & Blanchard, 2007; Noirot, 1972; Portfors, 2007; Portfors & Perkel, 2014; Sales, 1972; Scattoni, Crawley, & Ricceri, 2009; Takahashi, Kashino, & Hironaka, 2011; Takahashi, Thomas, & Barfield, 1983; Thomas, Takahashi, & Barfield, 1983). While analyses of emitted USVs have allowed researchers to study social and emotional dysfunctions—including affective disorders, addiction processes, and neurodevelopmental aberrations—perhaps the most interesting aspect of USV communication, particularly as studied in rats, is its role in emotional expression and motivation. Emotional behavior is one of the most difficult topics in biomedical studies; emission of vocalizations provides the best access to brain mechanisms governing this behavior.

II Significance of Studies of Emission of USVs

The study of the communication value of USVs has far-reaching clinical and practical applications by providing a window into processes that determine the course of development, emotional functions and dysfunctions, and the effects of environmental factors. As Rosenblatt and Lehrman (1963) so eloquently demonstrated many years ago in their studies of rat pups and dams, we can illuminate the processes underlying the development of behavior by analyzing the ways in which the current organization of behavior has arisen from preceding developmental experiences and then gives rise to succeeding ones. The temporal and acoustic organization of USVs provide a particularly sensitive window into this developmental process by characterizing a behavior that both reveals the effects of preceding experiences on the integrity of neurobehavioral organization and contributes to subsequent developmental stages of social behavior with salient vocal communicative value. Within this conceptual framework, we can underscore the processes and mechanisms that organize the production of temporal and acoustic characteristics of USVs; the preceding conditions that contribute to the development of variations in these temporal and acoustic characteristics; and the communicative and functional significance of these variations for the social environment.

Several chapters in this volume describe the processes and mechanisms that produce the temporal and acoustic characteristics of USVs in several infant and adult species. These processes include distal evolutionary contributions as well as proximal mechanisms, including laryngeal activity, peripheral motor dynamics, respiratory function, and ascending activating systems that change the state of the organism. To the extent that the action of proximal mechanisms is affected by changes in homeostasis or insults to the integrity of their neurobehavioral function, variations in USV characteristics are produced. As such, variations in USV characteristics may serve as an assessment of changes in arousal, emotional state, homeostasis, and changes in sociobiological conditions and environmental stressful events. In clinical application, these variations in vocal production may signal the threshold at which a wide range of adverse developmental conditions, such as exposures to toxic substances, pharmaceuticals, licit and illicit drugs, neuroactive agents, and various disease states, insult neurobehavioral integrity.

Further, to the extent that the neurophysiological basis of the mechanisms of USV production are understood, the diagnostic utility of USV analysis may be extended to uncovering the processes by which a given condition affects development or regressive processes in adulthood (e.g., autism and psychoses). Lastly, the high sensitivity of USVs to changes in neurobehavioral function accentuates the diagnostic utility of USV analyses. Variations in temporal and acoustic characteristics of USVs may be evident in the absence of evidence in gross assessment methods, including animal learning paradigms and measures of structural and/or physiological damage. Thus, studies of USVs may provide unique insight into brain processes and regulations, emotional states and motivations, and pathological processes that could not be observed otherwise.

Like other mammalian vocalizations, USVs can be conceptualized as a biologically significant ethotransmission, a behavior that not only is sensitive to changes in neurobehavioral organization as described earlier but also communicates the nature of that change to the social environment. Importantly, the USV rapidly communicates emotional states both to conspecifics and researchers. As such, we can examine how variations in the temporal and acoustic characteristics of USVs contribute to, and even create, the subsequent context within which the social environment responds. USV characteristics may elicit, for example, maternal responses that act to ameliorate developmental deficits or that act to exacerbate that developmental deficit through withdrawal or neglect. In the latter case, the lack of maternal responses to USVs may provide insight into such problems as the development of child abuse and/or neglect. Insights into how vocalizations contribute to the course of development are accentuated by translational evidence that supports more direct comparisons between findings of correlational studies of humans and findings of experimental studies of other species (Zeskind et al., 2014).

III Content of the Handbook

The present Handbook provides reviews of most major research aspects of emission of USVs and laboratory studies using USV measurement as a tool. Also, other vertebrate species that can emit USVs are described. The initial sections of the Handbook (Part A–C) deal with a brief history of bioacoustics, the principles of bioacoustic signal recording and analysis, aspects of the evolution of the social engagement system in mammals that is relevant to the development of USV, and laryngeal and respiratory mechanisms of USV production in rodents (Part C). Part D summarizes research on perceptual and auditory functions related to USVs, mechanisms of rapid transmission and initiation of emotional arousal (ethotransmission) in response to USVs, and their role in the development of social competence in young rats. In the following sections, USVs emitted by several rodent species, both as juveniles (Part E) and as adults (Part F), are described and acoustically characterized. They include juvenile rats, mice, and voles as well as adult rats, different species of mice, hamsters, gerbils, and voles. All those species were extensively studied as to their biology, ecology, and bioacoustics repertoire and are extensively used as model organisms.

Laboratory studies of emission of USVs as expression of emotional arousal and emotional states are discussed in the next two sections (Parts G and H), considering neurophysiological, neuroanatomical, neurobehavioral, and neuropharmacological mechanisms. Two ascending tegmental reticular systems responsible for the initiation of negative or positive emotional arousal and resulting vocal signaling are described. In rats, activation of one or the other of these systems will elicit emission of 22 kHz USVs or 50 kHz USVs, respectively. Thus, characteristics of rat-aversive 22 kHz USVs and appetitive 50 kHz USVs are provided, and interaction between emotional state and motivation is discussed. Emission of USVs may reflect the effects of numerous neuroactive substances, neurotransmitters, and drugs of abuse. Effects of psychostimulants, serotonin, cannabinoids, and morphine on the emission of USVs are described in detail in Part H.

The emission of USVs in rodents is a sensitive index of changes in many environmental, biological, and sociobiological conditions. The effects of physical stress and environmental and social effects as well as sexual differences in response to stress and effects of age on emission of USVs are provided in Part I. Sensitivity of emission of USVs to many external and internal factors, including expression of emotional states, made rodents particularly suitable for modeling and studying neuropsychiatric disorders and neurodevelopmental problems. Part J describes changes of USVs during cocaine self-administration, the effects of alcohol, and studies of schizophrenia. The next section describes the details of rodent research on autism and neurodevelopmental abnormalities as well as the effects of a mother's sleep deprivation on the offspring (Part K).

Finally, the emission of USVs or ultrasonic components of vocalizations of animals from other taxa are provided for comparison. They include emission of USVs in amphibians, bats, whales, and dolphins as well as selected species of small primates (Part L). Of particular importance are studies of vocal communication of aquatic animals and echolocation in bats.

The Handbook offers a unique compendium of extensive research endeavors and reveals the importance of studies of vocal communication in general as well as an insight into the brain mechanisms of behavior.

References

Barfield R.J., Thomas D.A. The role of ultrasonic vocalizations in the regulation of reproduction in rats. Annals of the New York Academy of Sciences. 1986;474:33–43.

Bass A.H., Gilland E.H., Baker R. Evolutionary origins for social vocalization in a vertebrate hindbrain-spinal compartment. Science. 2008;321(5887):417–421.

Brudzynski S.M. Communication of adult rats by ultrasonic vocalization: biological, sociobiological, and neuroscience approaches. ILAR Journal. 2009;50(1):43–50.

Brudzynski S.M. Social origin of vocal communication in rodents. In: Witzany G., ed. Biocommunication of animals. Dordrecht, The Netherlands: Springer Science + Business Media; 2014:63–79.

Kaltwasser M.-T. Acoustic signaling in the black rat (Rattus rattus). Journal of Comparative Psychology. 1990;104(3):227–232.

Litvin Y., Blanchard D.C., Blanchard R.J. Rat 22 kHz ultrasonic vocalizations as alarm cries. Behavioural Brain Research. 2007;182(2):166–172.

MacLean P.D. Brain evolution relating to family, play, and the separation call. Archives of General Psychiatry. 1985;42(4):405–417.

Noirot E. Ultrasounds and maternal behavior in small rodents. Developmental Psychobiology. 1972;5(4):371–387.

Portfors C.V. Types and functions of ultrasonic vocalizations in laboratory rats and mice. Journal of American Association for Laboratory Animal Science. 2007;46(1):28–34.

Portfors C.V., Perkel D.J. The role of ultrasonic vocalizations in mouse communication. Current Opinion in Neurobiology. 2014;28:115–120.

Rosenblatt J.S., Lehrman D.S. Maternal behavior of the laboratory rat. In: Rheingold H.L., ed. Maternal behavior in mammals. New York: Wiley; 1963.

Sales G.D. Ultrasound and aggressive behaviour in rats and other small mammals. Animal Behaviour. 1972;20(1):88–100.

Scattoni M.L., Crawley J., Ricceri L. Ultrasonic vocalizations: a tool for behavioural phenotyping of mouse models of neurodevelopmental disorders. Neuroscience and Biobehavioral Reviews. 2009;33(4):508–515.

Takahashi L.K., Thomas D.A., Barfield R.J. Analysis of ultrasonic vocalizations emitted by residents during aggressive encounters among rats (Rattus norvegicus). Journal of Comparative Psychology. 1983;97(3):207–212.

Takahashi N., Kashino M., Hironaka N. Structure of rat ultrasonic vocalizations and its relevance to behavior. PLoS One. 2011;5(11):e14115.

Thomas D.A., Takahashi L.K., Barfield R.J. Analysis of ultrasonic vocalizations emitted by intruders during aggressive encounters among rats (Rattus norvegicus). Journal of Comparative Psychology. 1983;97(3):201–206.

Zeskind P.S., McMurray M.S., Cox Lippard E.T., Grewen K.M., Garber K.A., Johns J.M. Translational analysis of effects of prenatal cocaine exposure on human infant cries and rat pup ultrasonic vocalizations. PLoS One. 2014;9(10):e10349.

Chapter 2

Ultrasonic Vocalizations, Their Recording, and Bioacoustic Analysis

Thomas MacDonald¹; Stefan M. Brudzynski²    ¹ Department of Technical Services, Brock University, St. Catharines, ON, Canada

² Department of Psychology, Brock University, St. Catharines, ON, Canada

Abstract

Studying ultrasonic vocalizations (USVs) requires appropriate recording and analysis systems that can manage high sound frequencies. One of the challenges of this approach is constructing a data acquisition system (DAS) that could process the frequency and dynamic range of the signals we wish to measure. This chapter will provide a brief history of measurements of ultrasonic signals and will explore the basic physical characteristics of mammalian USV parameters in time, frequency, power range, and spectrographic domains. An overview of the components of the DAS will be presented as will a description of how they are combined to make a fully functional recording and playback system for USV research. The most common digital signal-processing techniques will be examined while an explanation will be given as to how they can be used to improve recording and analysis techniques and assist in the study of vocalizations and their classification.

Keywords

Acoustic parameters; Data acquisition system; Digital recording and analysis of ultrasounds; Analog-to-digital conversion; Playback of ultrasounds; Microphones and loudspeakers; Sampling rate; Fast Fourier Transform

I Introduction

The discipline of bioacoustics was created at the very beginning of the 20st century through the research of Ivan J. Regen (1868–1947), who studied hearing, sound production, and the phonotaxis of insects (Orthoptera) (Regen, 1902, 1903). However, the term bioacoustics was coined a half-century later (c.1957) when this discipline was reborn. The increased interest in the subject happened due to technical developments that allowed for the conversion of sounds to electrical signals by sensitive microphones, amplification of these signals, visualization on oscilloscopes, and recording. However, the proper detection, recording, and analysis of ultrasounds did not become possible until after the development of so-called bat detectors.

The original idea of an ultrasound detector is attributed to George W. Pierce (1872–1956), one of the founding fathers of communication engineering (Brittain, 1997). The first ultrasound detector, which was a tuned superheterodyne detector, was originally developed by Noyes and Pierce (1938) and used in bioacoustics by Pierce himself to record ultrasound emitted by bush crickets (Gross, 2005; Pierce, 1948; Sales & Pye, 1974). Pierce's ultrasound detector was used as a bat detector for the first time in his own laboratory at the request of Donald Griffin (1915–2003), who was studying bats; thus, the first bat-generated ultrasounds were observed (Griffin, 1958; Gross, 2005; Pierce & Griffin, 1938; Sales & Pye, 1974). Griffin coined the term echolocation. Many fundamental observations on bat echolocation originated from that laboratory (Griffin & Galambos, 1941, 1942).

The first commercially produced bat detectors (broad-band envelope detectors), however, were not available until the mid-1960s (Halls, personal communication; Hooper, 1969; McCue & Bertolini, 1964). Under the direction of David Pye, a heterodyne-type bat detector with a broadband mode was built and developed to a tunable transistor version (Pye & Flinn, 1964; Sales & Pye, 1974) and eventually used in the QMC Mini and S100 bat detectors from Queen Mary College Instruments Ltd., founded in 1976. These bat detectors were suitable for use by both bioacoustics professionals and amateur naturalists.

The idea of providing broadband as well as tuned operation of the bat detector by a simple switch was realized in the 1970s. An analog bat detector was developed into the QMC S100 detector and later the QMC S200 detector with an added countdown receiver with envelope restoration, designed by Justin Halls (Halls, personal communication). This replaced the old broadband technique. The S200 bat detector had a simpler frequency division than the S100 with user-selected frequency division ratios of 4, 8, 16, and 32; it also provided the broadband option alongside the tuned option. These detectors were initially the most widely used in the research of animal ultrasound in the 1970s and 1980s and formed the foundation of modern ultrasonic bioacoustics.

Across the large diversity of bat detectors, the basic concept remained the same–to lower the ultrasonic sound frequency to an audible range for detection, observation, and analysis. For real-time recording, this was achieved through two main features: heterodyne modification and frequency division. In the heterodyne method, the original analog signal was modified by another tunable frequency and an audible signal was produced at a difference frequency. This method preserved the original signal amplitude although only a narrow band of the sound spectrum was scanned.

The frequency division method was done by a zero-crossing detector and the generation of a divided signal (square wave output every nth input signal) as an audible periodic signal. The frequency of the audible signal was related to the original input frequency by the user-selectable division factor (for further details, see Pye, 1980). This method reliably mapped a wide range of the sound spectrum, converting all detectable high-frequency signals to the audible range. The initial division technique, however, could not reproduce the amplitude of the original signal. Therefore, most bat detectors have a countdown function that mixes the output from the zero-crossing detector with an envelope detector so that the resulting output will preserve the amplitude of the incoming signals.

Modern technologies of bat-detector construction have been developed with new solutions such as digital frequency division, a time-expansion method, or some others digital sound-processing methods with the use of computer microcontrollers. In the time-expansion method, for example, the original signals can be recorded for a few seconds directly to an onboard digital memory, stored, and, after some delay, played back at a decreased speed. This system does not operate in real time so it does not preserve the original duration of the signal. However, it is sensitive and provides playback of the original, undivided signal. Most of the bioacoustic data cumulated in recent decades, however, originates from the two classical designs of bat detectors described earlier (Parsons, 1996; Waters & Walsh, 1994).

Although the detectors were gradually used to study sounds produced not only by bats but also by insects, rodents, and partly by birds, the traditional name bat detector has remained. Today, a large variety of manufacturers produce numerous types of bat detectors based on different principles and with improved capabilities, but mostly for field studies.

The advent of the digital signal processor (DSP) in the 1980s made a significant contribution to advancing the techniques available for recording and analyzing ultrasound. Prior to that, computers were too expensive and digital signal processing was limited to military, space, and medical applications (Smith, 1999). Many manufacturers (Texas Instruments, Motorola, Analog Devices) developed high-speed dedicated hardware processors in the mid-to-late 1980s, making it possible to record and process signals in real time using the mathematic algorithms of digital signal processing (Fourier transform, digital filtering, correlation, convolution). Today, this processing capability has progressed to the personal computer (PC), where the hardware is only required for the recording and playback of the signals with the data being transferred to and from the PC using a high speed USB port. All the DSP algorithms are now calculated on the PC and many bioacoustics software applications have an extensive library of DSP functions that can be used as effective tools for analyzing ultrasonic vocalizations (USVs).

This chapter will be mostly focused on the characteristics of the physical parameters of USVs and the description of the digital recording techniques.

II Physical Characteristics of USVs

The implementation of a USV recording and analysis system requires understanding the properties of the signal to be measured, namely the signal's general frequency bandwidth and dynamic range. The frequency bandwidth of sound is generally classified into four broad frequency ranges, shown in Table 2.1.

Table 2.1

USVs produced by animals, which are the main interest of this Handbook, are generated within frequencies of 20–100 kHz. Most vocalizations emitted by different species remain within the first half of that range (20–60 kHz), while only a limited number of species and their calls will be within the upper frequency range (60–100 kHz) with some even above 100 kHz.

USVs recorded from animals may be described in many ways. They may be labeled by their presumed biological function (e.g., alarm calls, distress calls), by an onomatopoeic label (i.e., a descriptor that verbally resembles the actual sound or an equivalent sound with lowered frequency, such as chirp or tsik calls), or even by the name that reflects the mechanism of their production (e.g., whistle vocalization). Because humans cannot hear ultrasounds, USVs are also often labeled by their sonographic appearance (e.g., flat call, sweep call, etc.). Other USV labels are more precise from the acoustic point of view and take into consideration their selected physical features (e.g., duration (long calls) or usual sound frequency, e.g., 50 kHz calls or 22 kHz calls). However, the most accurate description of vocalizations is by full physical characteristics of sound such as duration of calls (time domain), sound frequency and its structure (frequency domain), and the changes of sound frequency in time (spectrographic domain).

A Dynamic Range

The other important property used for sound classification is the dynamic range of sound. It is most often expressed in decibels (dB), a logarithmic unit that indicates ratio or gain of a signal to a reference. For illustration purposes, Table 2.2 shows the decibel range of common sounds within the human hearing range.

Table 2.2

An increase of 10 dB indicates an increase in the signal power level of 10 times the previous level. It is common to use the dB scale without using an absolute reference point but rather as a relative measurement from one sound to another. For example, a whisper can be + 10 dB greater that the rustle of leaves while at the same time − 40 dB less than a normal conversation. The decibel is the preferred unit of measure due to the large variance in power level over the human hearing range, as seen in Table 2.2. Because humans cannot perceive sounds higher than about 20 kHz, it is difficult to construct the relevant table for the ultrasonic range. One of the challenges of implementing a USV recording and analysis system is to select a data acquisition system (DAS) that meets the frequency and dynamic ranges of the signals we wish to measure.

B Time Domain and the Spectrogram

A graphical representation of sound is most often used to examine the characteristics of USVs where signals are commonly displayed in the time and frequency domains. A sample of a rodent USV is illustrated in Fig. 2.1 where the time window has been expanded to show each individual vocal component within that window.

Fig. 2.1 Time and spectrogram plot of a sample of a rodent 50 kHz call. The upper panel displays signal amplitude (in mV on the y -axis) against time (in seconds on the x -axis). The lower graph displays a spectrogram of the same signal, which is a plot of sound frequency (in kHz on the y -axis) against time (in seconds on the x -axis) against power (in decibels, color bar on the right side, z -axis).

The upper panel in Fig. 2.1 displays signal amplitude (in volts on the y-axis) against time (in seconds on the x-axis) while the lower panel displays a spectrogram of the same signal, which is a plot of sound frequency (in kHz on the y-axis) against time (in seconds on the x-axis). The signal amplitude is shown by color according to the color-coded scale (in dB on the right side). The spectrogram is the most widely used tool in bioacoustic analysis; it is used as a visual tool for examining and categorizing USVs by their appearance based on the frequency, power, and duration of each call element. In the upper amplitude/time graph, the voltage represents the USV signal recorded from the microphone, which has been digitized and recorded at fixed intervals along the time axis. Expanding, compressing, and shifting the time axis is a useful technique for visually searching and detecting USVs within long-duration recordings. Most software analysis programs allow the expanded window to be adjusted in time and to be shifted left or right to easily examine large data records. This technique resembles examining fragments of data with a magnifying glass.

The lower spectrogram plot in Fig. 2.1 is derived from time domain and is calculated using the discrete Fourier transform algorithm. The use of this method is beneficial for understanding the frequency range, power, and harmonic content of USVs. Fourier analysis is a mathematical technique that was named in honor of a French mathematician, Jean Baptiste Joseph Fourier (1768–1830). It states that any continuous periodic signal could be deconstructed as a sum of individual sinusoidal waves, each with its own amplitude and phase. The discrete Fourier transform (DFT) is one of many Fourier techniques that are used with digitized signals with fixed n data points. An interesting quality of the DFT is that it is a completely reversible transformation. The inverse discrete Fourier transform can reconstruct the original time domain of the signal from the frequency domain. This unique property is beneficial in USV analysis as it allows signals to be modified in either the time or frequency domains and then reconstructed (Smith, 1999). Amplitude and frequency resolution in both graphs in Fig. 2.1 are dependent on the analog-to-digital converter (ADC) and the sampling time interval selected, which are discussed in Section III B.

III Digital Recording Techniques

The major components of a USV DAS are illustrated in Fig. 2.2.

Fig. 2.2 Major components of a data acquisition system (DAS). Signal detections starts at the microphone where sound is converted to analog waveform. The signal is then passed to the analog amplifier/filter (A/F), then digitized to binary data with the ADC. Binary data is passed to the digital signal processor (DSP) for computational calculations. Correspondingly, binary data is converted to analog with the DAC and then passed on to the postamplifier/filter for signal conditioning to reconstruct the sound at the speaker.

Signal detection starts first with the microphone, where sound is converted to a continuous analog voltage waveform. This waveform is then passed through a preamplifier/filter to amplify the signal and to minimize noise and aliasing before being passed on to the ADC. The ADC converts the analog voltage to a binary number at successive discrete time intervals (for details, see Fig. 2.4) prior to being processed by the DSP/computer. The DSP/computer uses an extensive mathematical library of algorithms to process the measured signal both in the time and frequency domains.

The common analysis for USV signals includes their temporal characteristics, i.e., the duration of the signal, the time between acoustic components, and the number of USVs per unit time. It also includes the amplitude characteristics: peak amplitude and peak-to-peak amplitude. Analysis in the frequency spectrum includes peak frequency, minimum and maximum frequency, sound frequency bandwidth, and fundamental and harmonic frequencies. These analysis techniques allow us to characterize the structure of vocalization and its elements. The results of data collection and analyses are often passed on to a DAC to reconstruct the analog signal (Fig. 2.2, and for more details, see Fig. 2.4) for USV playback, if this is needed. The playback output signal can be used in experiments involving the responses of animal receivers to different acoustic stimuli; it can also be downshifted in frequency for the human audio hearing range. The DAC output signal is passed to a post amplifier/filter to condition the signal for speaker playback for listening (see Fig. 2.2). Each of the USV recording steps will be discussed in detail.

A Microphones and Preamplifier Filters

The two most common types of microphones used for USV recording are the prepolarized condenser microphones (electret) and the externally polarized condenser microphones (powered by 200 V). These microphones generally have low output voltage, typically less than 10 mV, and require a preamplifier to amplify the signal up to 1000 times (30 dB) prior to being sampled by the ADC. Prepolarized condensers are often used in portable equipment and are more useful in outdoor or high humidity environments. These microphones are low cost but have moderate sensitivity and relatively poor flat frequency response between 10 and 100 kHz (see Fig. 2.3, red plot). They are, however, still acceptable for many applications. The externally polarized condenser microphone is most often used in the laboratory environment and provides a good compromise between sensitivity and flat frequency response between 10 and 150 kHz (Fig. 2.3, blue plot). The externally polarized microphones are more costly due to an external 200 V power supply; they also require delicate handling due to the thin diaphragm inside.

Fig. 2.3 Frequency response (modified) of a prepolarized microphone (red trace, Knowles FG) and externally polarized microphone (blue trace, Avisoft CM16/CMPA). The external polarized microphone has an optimum flat frequency response up to 130 kHz while the prepolarized microphone starts to attenuate (roll off) in the ultrasonic range above 20 kHz. The y -axis shows signal attenuation in decibels and the x -axis sound frequency in kHz.

Ultrasonic signals are shorter in wavelength than audible sounds, so microphone placement for recording and soundproofing is important to reduce reflections and echoes. The best way to prevent this problem is to place the animal cage in a soundproof chamber lined with a material (foam, fabric) that absorbs ultrasound and prevents the signals from bouncing back to the microphone. For environments where a soundproofing chamber is not possible, measurements can be improved by adjusting the microphone height above the cage to maximize signal response for the cage area and adding a preamp filter to reduce interference below 10 kHz and above 100 kHz.

A common practice is to run background noise measurements prior to recording to determine additional possible sound sources. Overhead fluorescent lighting and adjacent computer hardware often generate frequencies in the ultrasonic band and can interfere with USV recording. Removing noise sources or repositioning equipment can greatly improve measurements. Special care should also be taken in a multicage environment to avoid cross talk between recording channels by using insulating materials between cages or positioning the cages and microphones at a distance that reduces interference. Postprocessing of signals by software can also be used to reduce unwanted noises that could be accidently recorded. This can be done by accepting only signals within selected frequency ranges and signal levels above a specified noise floor.

B Analog-to-Digital Conversion and Sampling Criteria

When selecting an ADC data acquisition board, the most important parameters are ADC bit resolution and sampling rate. Bits refer to digits of the binary code because computers record numbers as a series of ones and zeros. This bit count and sampling rate will determine the resolution, dynamic range, and frequency response we can expect for the desired USV measurements. Table 2.3 explains the effect of increased bit count (n) on the dynamic range.

Table 2.3

Calculations based on ADC/DAC convertor with 10 V input/output.

In general, increasing the number of ADC bits improves the dynamic range and resolution. An increase of two bits improves the voltage resolution by 4 and the dynamic range by 10 dB (10 ×). It should be noted that absolute resolution is dependent on the dynamic range of the microphone and the noise floor of the system. If the noise floor is above the bit resolution of the ADC, the resolution of the measurement is reduced (Higgins, 1990). In order to measure this accurately, more complex low noise amplifiers/filters and the soundproofing techniques discussed earlier would be required on the front end of the system to maintain the noise floor below the minimum voltage measurable by the ADC.

The sampling rate of the convertor will determine the maximum frequency the system can measure. The Nyquist (sampling) theorem states that the highest frequency that can be measured in a digitized analog system will equal one-half of the sampling rate (Smith, 1999). For USV measurements with a maximum frequency of 100 kHz, the absolute minimum sampling rate required would be 200 kHz. However, in practice it is best to oversample by 5–10 times, so ADCs with a sample rate of 500 kHz to 1 MHz are commonly used. It is also common practice to place a low-pass antialias filter prior to the analog-to-digital conversion (A/D) to remove any frequencies above the Nyquist frequency to prevent unwanted higher frequencies above that level to fold back as lower frequencies in the frequency spectrum (Smith, 1999).

Selection of the sampling rate may be illustrated for USVs of rats and mice. The category of 22 kHz USVs emitted by rats have frequencies from 19 to 32 kHz with a duration range from 300 to 4400 ms while the category of 50 kHz USVs has frequencies between 40 and 80 kHz with 15–100 ms call durations (Brudzynski, 2001, 2007). Based on the parameters of these USVs, an optimum ADC would be 16 bits/500 kHz providing a 96 dB range and a maximum sample rate of 500 kHz, which is roughly 10 times the 50 kHz call frequency. Research has shown that mice have an expanded frequency range of calls from 30 to 110 kHz (Portfors, 2007), which would indicate that an ADC with a sample rate of over 500 kHz would be better suited. Regardless of the ADC sampling rate, it is important to make sure that the selected microphone has the appropriate bandwidth and dynamic range required for the measurement. However, high sample rates place significant demands on data storage space (see Table 2.4) and a single channel sampled at 500 kHz requires 1.8 GB per hour or over 20 GB of space for an overnight run (12 h run).

Table 2.4

MB, megabyte; GB, gigabyte.

With terabyte drives available, the need for hard drive space may not seem like a significant issue. However, analyzing these large files can become time consuming and cumbersome where searching, for instance, for 100 ms call durations is analogous to finding a needle in a haystack. Luckily, most bioacoustic processing and analysis programs provide tools for quickly searching and locating USVs based on amplitude, frequency, and duration thresholds and allowing the bookmarking of areas of interest within these large files. Data compression can also take place in the acquisition phase by only allowing recording when the microphone exceeds an amplitude threshold of an active call. Applications of this type can record the live signal continuously to a circular buffer and, when a threshold trigger is detected, data recorded is stored on the hard drive at a preset time interval before and after a trigger. This is particularly useful for handling multiple channels in overnight runs where the animal calling is infrequent and storage capacity is limited.

C Digital-to-Analog Conversion for Playback

The playback of ultrasounds to the animal can be beneficial in projects when biological significance of the USVs is studied by animal behavioral responses to the signals. To accomplish this, we must reverse the process of analog-to-digital conversion (ADC) and convert the digital signal back to an analog signal using a digital-to-analog converter (DAC). In most commercially designed data acquisition boards, both the DAC and ADC are designed with the same bit width and sample rate, which makes the process of analog-to-digital and back to analog a seamless operation. The DAC with the same bit count and sample rate as the ADC would have the same dynamic range and frequency response. The output of the DAC must be conditioned prior to passing the signal on the post amplifier/speaker as the direct output of the DAC produces a staircase effect in the voltage due to the finite number of DAC codes (see Fig. 2.4). The staircase effect is an inherent property of the DAC as it produces only discrete output levels instead of a continuous analog signal. Minimizing the staircase effect can be improved by passing the DAC output to a low-pass analog filter with a cut-off frequency set just below the maximum frequency output (Smith, 1999). This amplifier filter is often part of the DAC component and, once the signal is conditioned, it can be passed on to the post amplifier/speaker.

Fig. 2.4 An illustration of the steps required to digitize and reconstruct an analog waveform. The continuous analog waveform is passed to the ADC where it is digitized to numeric amplitude values at discrete fixed time intervals. In order to reconstruct the waveform, the digital data is passed to a DAC at discrete time intervals and then passed to an analog low-pass filter to reconstruct the continuous analog waveform.

D Post Amplifier/Speaker

The output of the DAC has limited power capabilities so it must be connected to a post amplifier to enable it to drive a low impedance speaker load (4 Ω). Generally, an amplifier with a maximum power of 10 W will be sufficient to provide 70–90 dB dynamic range of sound at a distance of about a meter. The amplifier must have a wide-band frequency response (minimum 200 kHz) and may require a high-voltage output (200 V) if electrostatic speakers are used. The two most common types of speakers used for ultrasonic playback are dynamic and electrostatic types.

The dynamic speaker is often a good choice as it does not require external high voltage. However, care must be taken to ensure that the dynamic speaker has been designed for ultrasonic range. Standard audio tweeters have sharp frequency roll offs above the audio range, which will significantly decrease the power in the higher frequency range of USVs. Dynamic ultrasonic speakers generally give a flat frequency response up to 80 kHz before roll off is significant, which is adequate for many playback applications. For a wider frequency response above 100 kHz, the speaker of choice would be the ultrasonic electrostatic speaker, which can boast frequencies up to 150 kHz before significant frequency roll off begins.

A wideband sound level meter for ultrasound and a frequency generator are excellent tools to keep in the research laboratory to calibrate power levels for both microphones and amplifiers/speakers. Another useful feature in the playback of USVs is to concurrently downshift the signal into the human audible range by simply slowing down the audio playback rate from 200 kHz to computer soundcard frequency. This downshift will play back the sample at a slower rate, allowing the researcher to hear the audio output. This is a great method for detecting the presence of a USV without visual control. More complex techniques within a DSP mathematical tool kit involve demodulation and pitch shifting to bring the USV into the audible range if the playback rate is kept the same.

E DSP/Computer

The DSP/computer is the heart of USV data acquisition and analysis. It not only controls the data acquisition process but also performs USV analysis of the recorded data. These two functions are quite often separate entities. Data acquisition occurs in real time as signals are being measured while mathematical analysis is most often performed on stored data after the data acquisition stage (offline). The most common hardware configuration for USV recording is to use dedicated data acquisition hardware that connects directly to a computer USB port. USB high-speed performance eliminates the need to install dedicated hardware within the computer as data can be streamed in and out of the PC using PC-based hard disk recording software. This enhances the operational use for the USV recording system as data acquisition hardware can be connected to laptops, allowing for the portability of collecting data in field studies.

Data acquisition software for recording USV signals can often lead to confusion for the novice user as there are many control parameters that need to be set correctly for an effective measure of the signal of interest. The most common control parameters are listed in Table 2.5. We will now examine each of them in more detail as they relate to the digital recording in the time and frequency domains.

Table 2.5

1. Gain/Range Parameter

The channel gain/range parameter (Table 2.5) is used to set the input gain of the data acquisition hardware to maximize the dynamic range, which is the ratio of the largest to the smallest signal that can be measured. Care must be taken not to clip the signal with too much amplification, as shown in Fig. 2.5B. Clipped signals will add distortion and create false harmonics in the frequency domain, which can lead to poor detection and categorization of USV calls. Clipping can be removed by lowering the gain of the preamplifier or reducing the ADC voltage input range. Most ADC cards are programmable with voltage inputs gains from ± 100 mV to ± 10 V with increasing multiplications steps of 1, 2, 5, and 10.

Fig. 2.5 (A) Pure sinusoid waveform (left panel) generates a single frequency in the frequency spectrum (right panel). (B) A clipped sinusoid waveform generates harmonics in the frequency spectrum. (C) Incorrect ADC data formatting leads to erroneous higher frequencies in the frequency spectrum. Left panels show signal amplitude in V over time in μs and right panels show relative power in dB over frequency in kHz.

The A/D coding format parameter (Table 2.5) is specific to the ADC hardware module being used. There are many binary coding formats used with ADC, including binary, offset binary, and 2s complement integer (binary representation of signed integer numbers) with various bit widths, as listed in Table 2.3. The important issue is to have the data acquisition software matching the A/D coding format of the hardware so that correct measurements can be taken. Fig. 2.5C illustrates what happens when there is a mismatch between these settings. Recorded signals become distorted and the zero point of the signal switches from minimal to maximal values instantaneously. Most often this is not an issue with purchased recording software that works with one specific hardware module but many recording software programs allow various A/D modules with different bit resolutions and sample rates to optimize the data acquisition requirements. The software will provide control parameter(s) to match the ADC data format with the hardware