Source Microphone Research Articles

Source Microphone Identification Using Swin Transformer

Microphone identification is a crucial challenge in the field of digital audio forensics. The ability to accurately identify the type of microphone used to record a piece of audio can provide important information for forensic analysis and crime investigations. In recent years, transformer-based deep-learning models have been shown to be effective in many different tasks. This paper proposes a system based on a transformer for microphone identification based on recorded audio. Two types of experiments were conducted: one to identify the model of the microphones and another in which identical microphones were identified within the same model. Furthermore, extensive experiments were performed to study the effects of different input types and sub-band frequencies on system accuracy. The proposed system is evaluated on the Audio Forensic Dataset for Digital Multimedia Forensics (AF-DB). The experimental results demonstrate that our model achieves state-of-the-art accuracy for inter-model and intra-model microphone classification with 5-fold cross-validation.

Transformer for authenticating the source microphone in digital audio forensics

In recent years, digital audio has played a significant role in providing information used as evidence in criminal investigations. The authenticity of an audio recording must be proven for it to be admitted into evidence in a legal proceeding. In this paper, we propose a robust system based on transformer deep learning architecture for identifying the source microphone with high accuracy, which is applicable in digital audio forensics. To evaluate the proposed model, we use the Audio Forensic Dataset for Digital Multimedia Forensics (AF-DB) and the King Saud University speech database (KSU-DB) datasets. Experiments show that the proposed model achieves state-of-the-art classification accuracy on the two databases. The model achieves an overall accuracy of 94.18% and 84.19% for Inter-model and Intra-model microphone classification, respectively, for the AF-DB and 99.38% for the KSU-DB speech dataset.

Effect of Microphone Configuration and Sound Source Location on Speech Recognition for Adult Cochlear Implant Users with Current-Generation Sound Processors.

Microphone location has been shown to influence speech recognition with a microphone placed at the entrance to the ear canal yielding higher levels of speech recognition than top-of-the-pinna placement. Although this work is currently influencing cochlear implant programming practices, prior studies were completed with previous-generation microphone and sound processor technology. Consequently, the applicability of prior studies to current clinical practice is unclear. To investigate how microphone location (e.g., at the entrance to the ear canal, at the top of the pinna), speech-source location, and configuration (e.g., omnidirectional, directional) influence speech recognition for adult CI recipients with the latest in sound processor technology. Single-center prospective study using a within-subjects, repeated-measures design. Eleven experienced adult Advanced Bionics cochlear implant recipients (five bilateral, six bimodal) using a Naída CI Q90 sound processor were recruited for this study. Sentences were presented from a single loudspeaker at 65 dBA for source azimuths of 0°, 90°, or 270° with semidiffuse noise originating from the remaining loudspeakers in the R-SPACE array. Individualized signal-to-noise ratios were determined to obtain 50% correct in the unilateral cochlear implant condition with the signal at 0°. Performance was compared across the following microphone sources: T-Mic 2, integrated processor microphone (formerly behind-the-ear mic), processor microphone + T-Mic 2, and two types of beamforming: monaural, adaptive beamforming (UltraZoom) and binaural beamforming (StereoZoom). Repeated-measures analyses were completed for both speech recognition and microphone output for each microphone location and configuration as well as sound source location. A two-way analysis of variance using mic and azimuth as the independent variables and output for pink noise as the dependent variable was used to characterize the acoustic output characteristics of each microphone source. No significant differences in speech recognition across omnidirectional mic location at any source azimuth or listening condition were observed. Secondary findings were (1) omnidirectional microphone configurations afforded significantly higher speech recognition for conditions in which speech was directed to ± 90° (when compared with directional microphone configurations), (2) omnidirectional microphone output was significantly greater when the signal was presented off-axis, and (3) processor microphone output was significantly greater than T-Mic 2 when the sound originated from 0°, which contributed to better aided detection at 2 and 6 kHz with the processor microphone in this group. Unlike previous-generation microphones, we found no statistically significant effect of microphone location on speech recognition in noise from any source azimuth. Directional microphones significantly improved speech recognition in the most difficult listening environments.

Handclap for Acoustic Measurements: Optimal Application and Limitations

Handclap is a convenient and useful acoustic source. This study aimed to explore its optimal application and limitations for acoustic measurements as well for other possible utilizations. For this purpose, the following steps were performed: investigation of the optimal hand configuration for acoustic measurements and measurements at different microphone source distances and at different spaces and positions. All measurements were performed with a handclap and a dodecahedron speaker for comparison. The results indicate that the optimal hand configuration (among 11) is with the hands cupped and held at an angle due to the superior low frequency spectrum. This configuration produced usable acoustic parameter measurements in the low frequency range in common room background levels unlike other configurations. The reverberation time was measured across different spaces and positions with a deviation less than three and just a noticeable difference of the signal-to-noise ratio within or near the ISO 3382-1 limits for each corresponding octave band. Other acoustic parameters (i.e., early decay time, clarity) were measured with greater deviations for reasons discussed in the text. Finally, practical steps for measurements with a handclap as an acoustic source are suggested.

Utilizing Electrocochleography as a Microphone for Fully Implantable Cochlear Implants

Current cochlear implants (CIs) are semi-implantable devices with an externally worn sound processor that hosts the microphone and sound processor. A fully implantable device, however, would ultimately be desirable as it would be of great benefit to recipients. While some prototypes have been designed and used in a few select cases, one main stumbling block is the sound input. Specifically, subdermal implantable microphone technology has been poised with physiologic issues such as sound distortion and signal attenuation under the skin. Here we propose an alternative method that utilizes a physiologic response composed of an electrical field generated by the sensory cells of the inner ear to serve as a sound source microphone for fully implantable hearing technology such as CIs. Electrophysiological results obtained from 14 participants (adult and pediatric) document the feasibility of capturing speech properties within the electrocochleography (ECochG) response. Degradation of formant properties of the stimuli /da/ and /ba/ are evaluated across various degrees of hearing loss. Preliminary results suggest proof-of-concept of using the ECochG response as a microphone is feasible to capture vital properties of speech. However, further signal processing refinement is needed in addition to utilization of an intracochlear recording location to likely improve signal fidelity.

Source Microphone Recognition Aided by a Kernel-Based Projection Method

Microphone recognition aims at recognizing different microphones based on the recorded speeches. In the literature, Gaussian Supervector (GSV) has been used as the feature vector representing a speech recording, which is obtained by adapting a universal background model (UBM). However, it is not clear how the performance of the GSV will be affected by the number of mixture components in the UBM. Besides, the raw GSV obtained from a speech recording contains both the microphone response information and the speech information, meaning that the raw GSV can be quite noisy as the feature vector for microphone recognition. In this paper, we investigate how GSV will be affected by the UBM and other parameters during the calculation of the GSV. In addition, in order to improve the quality of the raw GSV, we propose a kernel-based projection method to be applied to the raw GSV. This projection method maps the raw GSV onto another dimensional space. It is expected that in the projected feature space, the microphone response information and the speech information can be separated into different dimensions, meaning that the projected GSV should be better as the feature vector for microphone recognition compared to the raw GSV. Two classifiers that have been used in the literature, namely linear support vector machine (SVM) and sparse representation-based classifier (SRC), are employed to compare the performance of the raw GSV and the projected GSV. The experimental results demonstrate that the projected GSV can outperform the raw GSV no matter using linear SVM or SRC as the classifier, which shows the effectiveness of the projection method.

Source microphone identification from speech recordings based on a Gaussian mixture model

Microphone identication is a specic type of media forensics that investigates whether it is possible to identify the source microphone from speech recordings. The main aim of this study is to nd out which of the several feature extraction techniques are best suited to the source microphone identication systems. We perform microphone identication experiments with 16 dierent microphones using 3 datasets. In order to improve the results on the datasets, we also investigate the important parameters that may aect the microphone identication performance. Our experimental results show that the proposed method is comparable to the existing studies in a closed-set identication

High intensity air ultrasound source for determining ultrasound microphone sensitivity up to 400 kHz

A broadband air jet ultrasound source, RSS101U-H, for animal bioacoustics research produces broadband ultrasound to at least 400 kHz. Free field reciprocity calibration of ½” condenser microphones on-axis, grid caps removed, using sine wave excitation by the method of Rudnick and Stein [JASA 20, pp 818-825, (1948)] was made (previously used to 280 kHz [JASA 67, p 7, (1980)]). Measurement from 5 kHz to 100 kHz was made in 1 kHz bins via an FFT analyzer. A communications receiver was used from 40 kHz to 400 kHz. The sensitivity of a ¼” microphone was determined from the free field of the reciprocity ½” microphone source. Air jet ultrasound level at 80 mm distance was then determined with the ¼“ microphone. Communications receiver 2.5 kHz bandwidth data was reduced to 1 kHz bin values. Air humidity sound absorption was determined via ANSI 1.26. The ¼” microphone sensitivity and broadband source sound level results in 1 kHz bands to 400 kHz are presented. Air jet spectral level was 97 dB re 20 uPa @ 75 kHz to 57 dB @ 400 kHz. This can be used to rapidly determine the sensitivity of any air ultrasound microphone over this frequency range.

Optimal microphone placement for active noise control in a forced-air cooling system.

In active noise cancellation systems with relatively small acoustic coupling, feedforward compensation is an effective methodology to create a controlled emission for sound attenuation. Especially for small electronic systems where forced air-cooling is required to control the temperature of large power sensitive components in the system, an active noise cancellation (ANC) system is a viable solution to reduce acoustic emissions. In this paper we discuss the placement of the noise source microphone for feedforward based active noise control in a forced-air cooling system. Noise source microphone placement is directed by the ANC performance of an on-line output-error based affine optimization of a linearly parametrized generalized finite impulse response filter for sound compensation. For the computation of the optimal filter, generalized or orthogonal FIR models are used as they exhibit the same linear parametrization as a standard FIR filter. The procedure is demonstrated on a small portable NEC LT170 data projector. The data projector is equipped with a shielded internal directional pick-up microphone to measure the sound created by the forced-air cooling of the projector’s light bulb. Non-invasive small directional speakers located at the inlet and outlet grill of the data projector are used to minimize acoustic coupling.

An Inexpensive Distance Learning Solution for Delivering High-Quality Live Broadcasts1

Providing an adequate method of distance learning is a challenge faced by many multicenter residency programs. The delivery of live didactics over the Internet is a convenient means of providing a uniform and equivalent educational experience to residents at distant sites. An application called MedCast has been developed with use of existing technologies, without the need for costly commercial products or equipment. MedCast captures the presenter’s computer screen and audio from a microphone source to produce a streaming video that is transmitted online and archived on a local server. Offsite residents can view broadcasts in real time or access archived conference sessions for later viewing. MedCast is available for download at no cost and offers several advantages, including a user-friendly graphical display interface, near-perfect preservation of image quality, and cost efficiency. Future plans include objective assessment of the efficacy of MedCast by comparing postlecture examinations to help evaluate for any differences between on- and offsite residents in terms of knowledge gained. A movie clip to supplement this article is available online at http://radiographics.rsna.org/content/28/5/1251/suppl/DC1. © RSNA, 2008

An Inexpensive Distance Learning Solution for Delivering High-Quality Live Broadcasts

Providing an adequate method of distance learning is a challenge faced by many multicenter residency programs. The delivery of live didactics over the Internet is a convenient means of providing a uniform and equivalent educational experience to residents at distant sites. An application called MedCast has been developed with use of existing technologies, without the need for costly commercial products or equipment. MedCast captures the presenter's computer screen and audio from a microphone source to produce a streaming video that is transmitted online and archived on a local server. Offsite residents can view broadcasts in real time or access archived conference sessions for later viewing. MedCast is available for download at no cost and offers several advantages, including a user-friendly graphical display interface, near-perfect preservation of image quality, and cost efficiency. Future plans include objective assessment of the efficacy of MedCast by comparing postlecture examinations to help evaluate for any differences between on- and offsite residents in terms of knowledge gained. A movie clip to supplement this article is available online at http://radiographics.rsnajnls.org/cgi/content/full/285085701/DC1.

Free listening‐point synthesizing method in a large microphone array using acoustic transfer function

In this research a free listening‐point synthesizing method is proposed by using the measured acoustic transfer function (ATF) in a large microphone array for future 3‐D audiovisual systems. ATFs at the location of far‐field microphones are estimated in frequency domain by using a recorded time‐stretched pulse (TSP) signal broadcasted by speakers at the location of sources. Then an ATF pool is generated with a chosen density by synthesizing the estimated ATFs for each sound source. The ATF pool contains all the microphones’ locations, plus synthesized ATFs with a chosen density from sound sources to any point on the arc which passes all of the microphones. The free listening‐point is generated by selecting the ATF for each source from the ATF pool. Then the source microphone sounds are convolved with the chosen ATFs and added. The proposed method is evaluated by an array with 38 microphones in the far field on an arc with r=4 m with 20 cm between each pair of microphones, and two microphones close to two sound sources with 1.5‐m distance. Experimental results have been shown for the estimated ATFs and comparison of the synthesized arbitrary listening point with the actually recorded sound for a given point that shows satisfactory results.

Free-field calibration of ‘‘1/2-in.’’ microphones

A precise free-field calibration system of type LS2 microphones by the reciprocity technique is proposed. A source and a receiver microphone are each attached to a cylinder of 1/2 in. diameter, respectively, and the cylinders are set up perpendicularly at the center of the anechoic room. The output voltage of the receiver microphone and the input voltage of the source microphone are measured by an FFT analyzer, and the frequency transfer functions are averaged to improve the signal-to-noise ratio. To reduce the influence of the disturbing reflections in the anechoic room, the gating technique is adopted. Values for the position of the acoustic centers of MR112 microphones were measured by two methods—gating and conventional. The uncertainty of the values by the gating method was less than by the conventional method.

Determination of acoustic center correction for type LS2aP condensor microphones

Acoustic center corrections for microphones are necessary for accurate free-field calibration by the reciprocity technique. Such corrections for several pairs of the relatively new type LS2aP microphone [IEC Std. 1094-1: 1992] were obtained by utilizing the theoretical inverse relationship between the sound pressure amplitude at the receiving microphone and the distance between the acoustic centers of the source and receiving microphones. For a nominally 10-V sinusoidal rms excitation of the source microphone, source-to receiver voltage ratios were measured with a dynamic signal analyzer at 500-Hz intervals in the extended frequency range (2 to 50) kHz. This procedure allowed all the data for a microphone pair to be gathered within several hours as a function of microphone diaphragm separation at 10-mm intervals from 101 to 311 mm. At each frequency, these ratios were corrected for atmospheric effects, including attenuation of sound, and then fit to a straight line (ratio versus diaphragm separation. Acoustic center corrections were calculated from the derived values of the fit parameters. These corrections agree with appropriately scaled values [IEC Publication 486-1974] for ‘‘1-in.’’ microphones with recessed diaphragms. Physical phenomena that cause small deviations from the linear fits will be discussed along with uncertainty estimates, and effects of spatially truncating data.

Noise considerations in capacitative source microphones for telephone applications

The signal-to-noise performance of two types of piezoelectric ceramic microphones designed for telephone applications is compared with that of two electret microphones. It is shown that for the microphone sensitivities considered the noise produced by low-cost preamplifiers suitable for low-voltage operation in a telephone set is the limiting factor. While the electret microphone-preamplifier combinations are some 23 dB more sensitive, they are also some 21 dB noisier than the ceramic-preamplifier types considered so that the resulting signal-to-noise ratios for the telephone bandwidth (C-message weighting) are comparable.