Headset Microphone Research Articles

Evidence-Based Recommendations for Tablet Recordings From the Bridge2AI-Voice Acoustic Experiments

BackgroundAs part of a larger goal to create best practices for voice data collection to fuel voice AI research, the objective of this study was to investigate the ability of readily available iOS and Android tablets with and without low-cost headset microphones to produce recordings and subsequent acoustic measures of voice comparable to “research quality” instrumentation. MethodsRecordings of 24 sustained vowel samples representing a wide range of typical and disordered voices were played via a head-and-torso model and recorded using a research quality standard microphone/preamplifier/audio interface. Acoustic measurements from the standard were compared to two popular tablets using their built-in microphones and with low-cost headset microphones at different distances from the mouth. ResultsVoice measurements obtained via tablets + headset microphones close to the mouth (2.5 and 5cm) strongly correlated (r’s > 0.90) with the research standard and resulted in no significant differences for measures of vocal frequency and perturbation. In contrast, voice measurements obtained using the tablets’ built-in microphones at typical reading distances (30 and 45cm) tended to show substantial variability in measurement, greater mean differences in voice measurements, and relatively poorer correlations vs. the standard. ConclusionFindings from this study support preliminary recommendations from the Bridge2AI-Voice Consortium recommending the use of smartphones paired with low-cost headset microphones as adequate methods of recording for large-scale voice data collection from a variety of clinical and non-clinical settings. Compared to recording using a tablet directly, a headset microphone controls for recording distance and reduces the effects of background noise, resulting in decreased variability in recording quality. Data Availability StatementData supporting the results reported in this article may be obtained upon request from the contact author.

Validity of Acoustic Measures Obtained Using Various Recording Methods Including Smartphones With and Without Headset Microphones.

The goal of this study was to assess various recording methods, including combinations of high- versus low-cost microphones, recording interfaces, and smartphones in terms of their ability to produce commonly used time- and spectral-based voice measurements. Twenty-four vowel samples representing a diversity of voice quality deviations and severities from a wide age range of male and female speakers were played via a head-and-thorax model and recorded using a high-cost, research standard GRAS 40AF (GRAS Sound & Vibration) microphone and amplification system. Additional recordings were made using various combinations of headset microphones (AKG C555 L [AKG Acoustics GmbH], Shure SM35-XLR [Shure Incorporated], AVID AE-36 [AVID Products, Inc.]) and audio interfaces (Focusrite Scarlett 2i2 [Focusrite Audio Engineering Ltd.] and PC, Focusrite and smartphone, smartphone via a TRRS adapter), as well as smartphones direct (Apple iPhone 13 Pro, Google Pixel 6) using their built-in microphones. The effect of background noise from four different room conditions was also evaluated. Vowel samples were analyzed for measures of fundamental frequency, perturbation, cepstral peak prominence, and spectral tilt (low vs. high spectral ratio). Results show that a wide variety of recording methods, including smartphones with and without a low-cost headset microphone, can effectively track the wide range of acoustic characteristics in a diverse set of typical and disordered voice samples. Although significant differences in acoustic measures of voice may be observed, the presence of extremely strong correlations (rs > .90) with the recording standard implies a strong linear relationship between the results of different methods that may be used to predict and adjust any observed differences in measurement results. Because handheld smartphone distance and positioning may be highly variable when used in actual clinical recording situations, smartphone + a low-cost headset microphone is recommended as an affordable recording method that controls mouth-to-microphone distance and positioning and allows both hands to be available for manipulation of the smartphone device.

The effects of delay on task-based interactive conversation

Previous studies have found that, in the presence of noise, the distribution of floor transfer offsets (FTO) broadens and shifts to the right, and it has been suggested that this is due to increased listening effort. The present study investigates if speakers are sensitive to the timing of the start of a partner's turn during interactive conversation. By manipulating the delay on the channel between two talkers on a turn-by-turn basis, the floor transfer offset (FTO) distribution can be modified to simulate what has been observed in more challenging conversational settings. Talkers were seated in different rooms but could communicate via headset microphones and headphones with gains fixed such that the levels would simulate the two talkers sitting in the same room. Conversations where pairs of talkers solved the DiapixUK spot-the-difference task were conducted in both quiet and noise, both with and without added delay. Measures of speech production (e.g., speech level, articulation rate, etc.) and conversation behavior (FTO, turn length, etc.) are compared across the conditions.

The effect of task on speech production and conversation behavior during conversation

The purpose of this study was to investigate how native-English, healthy-hearing individuals adapt their speech production and conversation behavior in the presence of noise and how this can vary based on conversational goal. Pairs of participants engaged in both free-form conversations as well as conversations based on solving a task (a “spot the difference” task using the Diapix UK pictures). Although seated in separate rooms, talkers could communicate via headset microphones and headphones with gains set to simulate levels that would be present if they were seated in the same room. The effects of task and noise on measures of speech production (e.g., articulation rate, speech level, etc.) and conversational behaviors (e.g. floor transfer offsets, turn length, etc.) are investigated. These results provide insights into how to infer listening effort via acoustical measures of communication in a broader range of settings.

Conversation behavior assessment to estimate communication difficulty

The communication behavior of humans adapts to the needs of different communication situations. Humans have developed several strategies to successfully communicate in challenging acoustic environments. Some of these strategies can be consciously controlled while in a conversation, but others happen sub-consciously, or as a mixture of both. The applied strategies come with an increased effort, a deviation from common behavior, or the overcoming of personal comfort zones and are usually ceased as soon as the situation allows. This study’s aim is to monitor participant’s communication behavior in easy and challenging acoustic situations and use this data as a measure for the experienced communication difficulty. The Sonova Real Life Lab allows to investigate natural conversation situations in a controlled acoustic environment. Participants can move around and interact with each other across a 25m² stage, while their head positions and orientations are tracked with a motion capturing system. The voices of the participants are recorded using wireless headset microphones. This allows to estimate individual vocal effort and analyze backchanneling and turn taking behavior. This contribution will present and discuss data collected with participants engaged in real one-to-one conversations, while the acoustic background was systematically altered.

Background noise and reverberation time in a college classroom

A sound survey of noise levels in unoccupied classrooms on the Marietta Campus of Kennesaw State University identified one classroom as the most problematic in terms of background noise levels. This level is mostly associated with heating, ventilation, and air-conditioning equipment. Excessive background noise in learning spaces diminishes the speech intelligibility from the speaker to the listeners and can cause stress and fatigue. The overall A-weighted average sound levels are compared with ANSI-ASA S12.60-2010/Part 1 (Acoustical Performance Criteria, Design Requirements, and Guidelines for Schools). In addition, frequency analysis of the background sound and time variations are studies. Binaural headset microphone measurements are compared to the single microphone measurements. Reverberation time is measured using the engineering survey method and compared to the ANSI guidelines.

Two efforts towards natural speech synthesis: Incorporating disfluency and speaking style change based on the interlocutor

During the period 2000–2005, a Japanese female speaker recorded her everyday conversations with many different interlocutors using a head-set microphone. As a result, 600 hours of natural Japanese speech data were obtained. This study describes a DNN-based speech synthesis system which was trained on 300 hours of the data, focusing on two unique efforts to make it more expressive in a human-like way: (1) allowing for disfluencies, and (2) accounting for the category of interlocutor. Incorporating some frequently observed disfluent patterns in general Japanese speech such as fillers, phrase-final rising intonation, and word-internal prolongation or suspension, is believed to be effective in practical application as certain disfluencies are connected to a speaker’s attitude in Japanese communication. For example, having word-internal prolongations can show hesitation or politeness, and word-internal suspending can show the speaker’s surprised attitude. Interlocutors in the original data were categorized into four groups: family, friend, child and others. This information was used in the training process, and as a result, the synthesizer can generate different speaking styles according to the interlocutor setting. Being able to generate disfluent speech and change the speaking style depending on who you are talking to can make the synthesizer ever more expressive.

Normative Values of Cepstral Peak Prominence Measures in Typical Speakers by Sex, Speech Stimuli, and Software Type Across the Life Span.

The purpose of this study was to determine normative values for cepstral peak prominence measures across the life span as a function of sex using clinically relevant stimuli (/ɑ/, /i/, and two sentences of The Rainbow Passage) and two commonly used software types: Praat (Version 6.0.50) and Analysis of Dysphonia in Speech and Voice (ADSV). One hundred fifty speakers (75 males, 75 females; evenly distributed into three age groups) without voice disorders aged 18-91 years were recorded via headset microphone in a sound-treated booth. Cepstral measures were analyzed using common analysis methods in Praat and ADSV by sex, stimuli, and software type. Kruskal-Wallis tests and post hoc Mood's Median tests for significant factors were performed on cepstral measures to assess the effects of age group, sex, stimuli, and software type. The results revealed statistically significant effects of sex, stimuli, and software type on cepstral measures, but no statistical effect of age group on cepstral values. Females had lower average cepstral values compared to males. Across stimuli, the highest average cepstral measure was found for sustained /ɑ/, followed by sustained /i/, and then of the two sentences of The Rainbow Passage. Average cepstral measures in Praat were higher than those from ADSV. The current work did not find a statistical effect of age group on cepstral values; thus, normative cepstral values were reported by sex, stimuli, and software type. Future work should examine the applicability of these normative values for discriminating speakers with and without voice disorders.

Group conversation assessment in realistic acoustic scenes

Successful communication in background noise is a continuous interaction between conversation partners. They continuously exchange conscious, unconscious, verbal, and non-verbal information regarding how well the other person is understood. Human communication is not limited to one-on-one interactions but can also happen in group conversation situations. The complexity of taking an active part in a group conversation is much higher though, since the continuous interaction happens with more than one conversation partner at the same time. To investigate human communication behavior, real conversations between groups of six people were observed in the Sonova Real Life Lab. All participants were allowed to move around freely within the lab space. The head position and orientation of all six participants was continuously monitored with an optical motion capturing system. The voices of all participants were recorded with wireless headset microphones, which allowed to assess individual vocal effort. During the conversation situation, background noise was played from all four directions simultaneously and systematically varied in overall level. The individual position data and vocal effort were used to estimate perceived communication difficulty and signal to noise ratios. The head orientation data was analyzed to estimate which individuals the participants were talking and listening to.

Developing a voice monitoring smartphone app: Acoustic acquisition and processing considerations

With the advent of smartphone technology, there has been an increase in at-home health monitoring. Yet, there are few applications (apps) available to track voice acoustics for those with voice disorders. Therefore, we completed two investigations into the acquisition and processing of the acoustic signal to help develop a voice monitoring app. Study 1: We investigated how microphone distance and phone tilt impact the accuracy of acoustic measures of voice. A total of 58 participants with (n = 47) and without (n = 11) voice disorders completed speech tasks of sustained vowels and the rainbow passage. Concurrent recordings were collected using the participant’s phone, as well as a stationary headset microphone for comparison. We determined that the voice measures of fundamental frequency (Hz), voicing duration (seconds), and cepstral peak prominence (dB) were impervious to phone distance and tilt. Study 2: Focusing on the three acoustic measures from study 1, we investigated the correspondence between acoustic measures attained with clinically available software (e.g., Praat) and those from our own lab-developed algorithms specialized for on-app processing. Preliminary results show strong relationships (r > .90) between the different processing techniques. Further work is needed to understand how participant-specific factors (age, dysphonia severity) may improve algorithm accuracy.

Communication behavior observation

Communication is one of the most important aspects of human society. Humans have developed several strategies to successfully communicate in challenging acoustic environments. Some of these strategies can be consciously controlled while in a conversation, but others happen unconsciously or as a mixture of both. To investigate communication behavior, real two-person conversations were observed in the Sonova Real Life Lab. Within the lab, participants were allowed to move around freely in an acoustic scene. The head position and orientation of both participants were monitored with a motion capturing system and the voices of both participants were recorded with wireless headset microphones. The voice intensity and spectral content were analyzed from the speech recordings to assess vocal effort and detect Lombard speech. The audio setup of the Real Life Lab was used to playback an acoustic background scene, which could be directly manipulated in level and spectral content. This contribution shows data from pilot experiments in which the participants had to communicate in acoustically controlled background scenes. The data show the characteristics of and interactions between different strategies humans use to successfully communicate in challenging acoustic environments. Learnings for future experiments are discussed.

The Effect of Microphone Frequency Response on Spectral and Cepstral Measures of Voice: An Examination of Low-Cost Electret Headset Microphones.

The purpose of this study was to establish the frequency response of a selection of low-cost headset microphones that could be given to subjects for remote voice recordings and to examine the effect of microphone type and frequency response on key acoustic measures related to voice quality obtained from speech and vowel samples. The frequency responses of three low-cost headset microphones were evaluated using pink noise generated via a head-and-torso model. Each of the headset microphones was then used to record a series of speech and vowel samples prerecorded from 24 speakers who represented a diversity of sex, age, fundamental frequency (F o), and voice quality types. Recordings were later analyzed for the following measures: smoothed cepstral peak prominence (CPP; dB), low versus high spectral ratio (L/H ratio; dB), CPP F o (Hz), and cepstral spectral index of dysphonia (CSID). The frequency response of the microphones under test was observed to have nonsignificant effects on measures of the CPP and CPP F o, significant effects on the CSID in speech contexts, and strong and significant effects on the measure of spectral tilt (L/H ratio). However, the correlations between the various headset microphones and a reference precision microphone were excellent (rs > .90). The headset microphones under test all showed the capability to track a wide range of diversity in the voice signal. Though the use of higher quality microphones that have demonstrated specifications is recommended for typical research and clinical purposes, low-cost electret microphones may be used to provide valid measures of voice, specifically when the same microphone and signal chain is used for the evaluation of pre- versus posttreatment change or intergroup comparisons.

Skin Acceleration Levels Estimated by a Neck-surface Accelerometer during Phonation Are Affected by The Mechanical Properties of The Anterior Cervical Skin

The aim of the present study was to assess whether skin acceleration levels (SAL) estimated by a neck-surface accelerometer (ACC) are affected by the anterior neck skin condition. Each of six healthy non-dysphonic participants wore a headset microphone (MIC), had an ACC sensor fitted on the skin over the cervical trachea, and were subsequently asked to gradually increase the vocal loudness during sustained phonation of the vowel /e:/ (crescendo task), while the sound pressure levels on a sound level meter (SPLSLM) and MIC/ACC signals were simultaneously recorded. Root mean squared values were calculated from the MIC and ACC signals as the sound pressure level (SPLMIC) and SAL, respectively, and the relationships between SPLSLM and SAL were compared between neck anteflexion and retroflexion or between outward and inward skin retraction on both sides of the sensor. In the total samples for the successful crescendo performance in a natural head position, the SPLMIC and SPLSLM showed a strong linear correlation (r=0.980), whereas the correlation between the SAL and SPLSLM showed a distorted regression line (r=0.765) with individual differences. In all participants, the anteflexion and inward skin retraction decreased the SAL value at the same SPLSLM value, whereas the retroflexion increased the SAL value at the same SPLSLM value. These results demonstrate that the signal intensity of a neck-surface ACC is affected by the condition of the anterior cervical skin, perhaps leading to inter-individual variability in SAL measurements.

Musical improvisation enhances interpersonal coordination in subsequent conversation: Motor and speech evidence.

This study explored the effects of musical improvisation between dyads of same-sex strangers on subsequent behavioural alignment. Participants–all non-musicians–conversed before and after either improvising music together (Musical Improvisation—MI—group) or doing a motoric non-rhythmic cooperative task (building a tower together using wooden blocks; the Hands-Busy—HB—group). Conversations were free, but initially guided by an adaptation of the Fast Friends Questionnaire for inducing talk among students who are strangers and meeting for the first time. Throughout, participants’ motion was recorded with an optical motion-capture system (Mocap) and analysed in terms of speed cross-correlations. Their conversations were also recorded on separate channels using headset microphones and were analysed in terms of the periodicity displayed by rhythmic peaks in the turn transitions across question and answer pairs (Q+A pairs). Compared with their first conversations, the MI group in the second conversations showed: (a) a very rapid, partially simultaneous anatomical coordination between 0 and 0.4 s; (b) delayed mirror motoric coordination between 0.8 and 1.5 s; and (c) a higher proportion of Periodic Q+A pairs. In contrast, the HB group’s motoric coordination changed slightly in timing but not in degree of coordination between the first and second conversations, and there was no significant change in the proportion of periodic Q+A pairs they produced. These results show a convergent effect of prior musical interaction on joint body movement and use of shared periodicity across speech turn-transitions in conversations, suggesting that interaction in music and speech may be mediated by common processes.

Akustinen äänenlaatuindeksi (AVQI) äänen arvioinnissa: Alustava monitapaustutkimus äänitystason, -tilan ja äänentuottotavan vaikutuksista

Kuudesta parametrista muodostuvaa akustista äänenlaatuindeksiä (AVQI) käytetään äänihäiriöiden objektiiviseen arviointiin. Tämä tutkimus on alustava selvitys siitä, paljonko AVQI:in vaikuttavat äänitystaso, ympäristön taustakohina, puhujan puhetapa ja sukupuoli silloin, kun käytetään suositeltua äänityslaitteistoa.
 Näytteet tallennettiin kahdelta terveääniseltä (mies ja nainen) sekä äänihuulihalvauksesta (mies) ja toiminnallisesta äänihäiriöstä (nainen) kärsineeltä henkilöltä. Terveäänisten puhujien näytteet äänitettiin vaimennetussa studiossa. He lukivat lyhyen tekstin ja äänsivät [a:]-vokaalia 5 sekuntia itselleen tavanomaisella, sekä hiljaisella, voimakkaalla ja korkealla äänellä. Huonon äänitystason ja -ympäristön vaikutusten tarkastelua varten ääninäytteet soitettiin yksitiekaiuttimen kautta ja äänitettiin uudelleen studiossa (taustakohinataso 17,3 dBA) ja toimisto-olosuhteissa (taustakohinataso 34,4 dBA) kahdella äänitystasolla. Kaiuttimen kautta soitetuista näytteistä laskettiin AVQI.
 Tulosten mukaan äänitystaso, -tila ja sukupuoli vaikuttivat vain vähän AVQI-lukuun ja sen parametreihin. Eniten AVQI-lukua nosti häiriöinen äänenlaatu. Myös puhetapa vaikutti: hiljaisesti tuotetuissa ääninäytteissä terveääniset saivat äänihäiriöille tyypillisiä arvoja. AVQI:a käytettäessä on tärkeää kontrolloida, että puhujat tuottavat ääninäytteet itselleen tavanomaisella puhevoimakkuudella ja -korkeudella. Tämä on oleellista etenkin toistomittauksissa.

The implementation of Fast Hilbert Transform with low pass filter for reconstructing breathing sound signals in LabView

Many studies have stated that the respiration rate is one of the vital signs in individual health. It can be used to identify serious illness. Detecting either the chest movement or the exhausted gases properties using various sensors are the common methods to monitor respiration rate. Apart from both, breathing sound seems also effective for such purpose without requiring any complicated sensing devices. However, attaining respiration rate from the burst-type of breathing sound signals is challenging. Their high variations in frequency, amplitude, and the interval between inhalation and exhalation make the reconstruction of a smoother signal needs extra signals processing effort. This paper proposes a technique to reconstruct breathing sound which is acquired using a built-in headset microphone and implemented using Fast Hilbert Transform with Low Pass Filter in LabView. Through adjusting a proper filter order and cut-off frequency, the result shows that the proposed method is able to derive the representation of breathing sound signals in a smooth curve and convenient form. Compare to the raw ones, this reconstructed signals will make the peak detection easier. It is expected that this sound-based approach will be a cheap and effective solution for further respiration rate monitoring systems.

Skype & Type

Voice-over-IP (VoIP) software are among the most widely spread and pervasive software, counting millions of monthly users. However, we argue that people ignore the drawbacks of transmitting information along with their voice, such as keystroke sounds—as such sound can reveal what someone is typing on a keyboard. In this article, we present and assess a new keyboard acoustic eavesdropping attack that involves VoIP, called Skype & Type ( S&T ). Unlike previous attacks, S&T assumes a weak adversary model that is very practical in many real-world settings. Indeed, S&T is very feasible, as it does not require (i) the attacker to be physically close to the victim (either in person or with a recording device) and (ii) precise profiling of the victim’s typing style and keyboard; moreover, it can work with a very small amount of leaked keystrokes. We observe that leakage of keystrokes during a VoIP call is likely, as people often “multi-task” during such calls. As expected, VoIP software acquires and faithfully transmits all sounds, including emanations of pressed keystrokes, which can include passwords and other sensitive information. We show that one very popular VoIP software (Skype) conveys enough audio information to reconstruct the victim’s input—keystrokes typed on the remote keyboard. Our results demonstrate that, given some knowledge on the victim’s typing style and keyboard model, the attacker attains top-5 accuracy of 91.7% in guessing a random key pressed by the victim. This work extends previous results on S&T , demonstrating that our attack is effective with many different recording devices (such as laptop microphones, headset microphones, and smartphones located in proximity of the target keyboard), diverse typing styles and speed, and is particularly threatening when the victim is typing in a known language.

Do Smartphone Hearing Aid Apps Work?

The National Institute of Deafness and Communication Disorders of the National Institutes of Health under Award Number 5R21DC016241-02.

Acoustic Voice Quality Index as a Potential Tool for Voice Screening

To diminish the risk of voice disorders in people who are highly dependent on their voices, such as teachers, vocal screening is important already at the beginning of such individuals' professional studies. A reliable, specified screening tool is needed. The Acoustic Voice Quality Index (AVQI) has been found to differentiate normal voices from abnormal voices and to serve as a treatment outcome measure. This study investigated whether AVQI could be a screening tool in combination with auditory- and self-perception of the voice to discriminate normal from slightly poor voices. Experimental. Some 128 female teaching students (mean age 26.39 years, SD 9.80 years) with no diagnosed voice disorders participated in this study. They read aloud a text in Finnish, sustained the vowel /a:/, and filled the Voice Handicap Index (VHI) questionnaire. Voice samples were recorded with an AKG C544L headset microphone, iFocusrite soundcard, and Praat software using a 44100 sample rate and 16-bit amplitude quantization. Five expert voice therapists evaluated the samples to determine the grade of dysphonia (G) using a scale of 0-0.5 (=normal), 0.5-1 (=mild), 1-2 (=moderate), and 2-3 (=severe). Three medial seconds of [a:] and the first 31 syllables of the text were analyzed using AVQI script version 03.01 in Praat (5.3.55). The analysis gives one AVQI score per participant (scale 0-10). The AVQI threshold of normal and disordered voices for Finnish speakers is 1.83; a Gmean = 0.0-0.5 and VHI score <19 were considered normal. Statistical analysis was done using the receiver operating characteristic (ROC) curve, Spearman's correlation coefficient, and the independent samples t test. According to the AVQI results, the area under the curve (AROC) was 0.554, which is fair. The Youden index gave a cutoff value of 0.30 with a sensitivity of 85% and a specificity of 81.1%. There were weak but significant correlations between Gmean and AVQI and two AVQI parameters, smoothed cepstral peak prominence and harmonic-to-noise ratio (r = 0.27; --0.24; -0.20, respectively; |P < 0.05); and between total VHI and AVQI score and cepstral peak prominence (r = 0.21; 0.20, respectively; P < 0.05). Furthermore, the AVQI scores differed significantly between the groups with a VHI total score <19 and ≥19. AVQI did not differentiate between voices that had been perceptually judged as normal or slightly abnormal, but a combination of perceptual assessment in the form of AVQI and VHI could better screen slightly deviant voices.

Evaluations of the voice to text transfer in different conditions

The purpose of the research was evaluations of voice to text transfer in different conditons for the use in the verbal control of industrial robots. The research aims to find the best conditions to command an industrial robot using the human voice. A comprehensive study of existing problems and suggestions for simulation problems has been performed. Unlike some other works, it focused on the simply external problems to affect the voice to text transfer. The main problems of voice to text transfer (speech speed, speaker distance, ambient noise) has been established. The simulation using a personal computer equipped with a sound board and a headset microphone has been performed in all combinations of conditions. On the basis of the analyzes was establish the result of this research, the most suitable conditions and the worst conditions for voice to text transfer.