Running Autocorrelation Function Research Articles

Speaker recognition analysis using running autocorrelation function parameters

The human brain's process associated with the recognition and identification of acoustic signals is regarded as the calculation of the distances between “sound vectors;“ the ones listened in the present with the memorized vectors in previous listening. Matching vectors (minimal distance between them) would indicate sounds come from similar sound sources or same sound source. In this study, the statistical values of r-ACF (running autocorrelation function) microscopic parameters of 10 spoken words recordings (in Spanish) by the same and different speakers were calculated, and the aforementioned vector's distances were constructed by using the distances between the r-ACF parameters, being able to predict the degree of similarity between the speakers.

Investigation of calculation methods of effective duration on autocorrelation-based Chinese speech recognition

The factors extracted from running autocorrelation function have phonetic and syllabic meanings, and they can be used for Chinese speech recognition. One of the factors is the effective duration of ACF envelope τe, which reveals the temporal repetitive continuity of the speech signals. The values of τe vary when different calculation methods are adopted. By adopting different calculation methods to several sets of Chinese vowels and characters, which are recorded from a group of speakers, a preferred calculation method is decided by the uniqueness of τe, that is the τe calculated by such method can mostly distinguish different vowels, characters, or even speakers.

Synthesis of the speech signals by using autocorrelation function

The running autocorrelation function (r-ACF) is obtained by the FFT method based on the Wiener–Khinchine theorem after obtaining the power density spectrum for a signal. This study attempted to reconstruct the original speech signal by using a part of its r-ACF. First, the stationary part of the vowel signals were investigated to determine until which delay time of the ACF (maximum time lag) is necessary to recognize the reconstructed signals as the original ones. Then, the continuous speech signals were investigated to determine the appropriate integration interval as well as the maximum time lag.

Comparison of Different Calculation Methods of Effective Duration (<I>τ<SUB>e</SUB></I>) of the Running Autocorrelation Function of Music Signals

Comparison of Different Calculation Methods of Effective Duration (<I>τ<SUB>e</SUB></I>) of the Running Autocorrelation Function of Music Signals

Computer software for identification of noise source and automatic noise measurement

A new computational system for the environmental noise measurement and analysis has been developed. The system consists of binaural microphones, a laptop PC, and analysing software. A target noise is recorded automatically depending on the specified background noise level, and the acoustical parameters are calculated simultaneously. These functions allow for precise field measurements. The system is equipped with a template-matching algorithm for the identification of noise source. This function was implemented to avoid the effect of an interrupting sound such as voice and wind blowing during a measurement. Noise analyses in this system are based on the model of human auditory system. In addition to the time-series data of sound level, the important acoustical parameters of noise source are extracted from the running autocorrelation function (ACF) and the inter-aural cross-correlation function (IACF). It has been found that those parameters are strongly related to the auditory primary sensations and spatial sensations. Evaluation of the environmental noise based on these functions is another feature of this system. This paper describes the effectiveness of the ACF and the IACF analysis for analysing acoustical properties of noise and for evaluating the subjective response to noise.

Blending vocal music with a given sound field due to the characteristics of the running autocorrelation function of singing voices

This is a study to meet music and the opera house acoustics. It is said that singers adjust their interpretation style according to the acoustical condition of the sound field in a room. However, this mechanism of blending of musical performance with the sound field is unknown. In order to obtain a method of performance blending of opera house acoustics, we attempted to develop evaluation criteria for a singing voice in terms of the minimum value of the effective duration of the running autocorrelation function (r-ACF), (te)min, of sound source signals. This temporal factor has shown to have close correlation with the subjective response of both listeners and performers to sound fields [Y. Ando, Architectural Acoustics (AIP Press/Springer-Verlag, New York, 1998)]. As example for the control of (te)min due to performing style, effects of singing style, kind of vowel, relative pitch, vibrato extent, and intonation on the values of (te)min are demonstrated. In addition, the fine structure of the r-ACF is discussed with regard to the identification of vowels of singing voice. a)Now at 1-10-27 Yamanokami, Kumamoto, Japan.

Theory of acoustic design of opera house and a design proposal

First of all, the theory of subjective preference for sound fields based on the model of auditory-brain system is briefly mentioned. It consists of the temporal factors and spatial factors associated with the left and right cerebral hemispheres, respectively. The temporal criteria are the initial time delay gap between the direct sound and the first Reflection (Dt1) and the subsequent reverberation time (Tsub). These preferred conditions are related to the minimum value of effective duration of the running autocorrelation function of source signals (te)min. The spatial criteria are binaural listening level (LL) and the IACC, which may be extracted from the interaural crosscorrelation function. In the opera house, there are two different kind of sound sources, i.e., the vocal source of relatively short values of (te)min in the stage and the orchestra music of long values of (te)min in the pit. For these sources, a proposal is made here.

MELISMA SINGING AND PREFERRED STAGE ACOUSTICS FOR SINGERS

Vocal performance characteristics, such as tempo, colour, and expressiveness of singers are part of the unique artistic impression of individual performers and individual performances. Subjective pair-comparison studies of singers in a previous study demonstrated that singers prefer added reflections with delays in the range of 10–20 ms. However, the range of values between singers, for the effective duration of the autocorrelation of the singer's voice, was limited, and insufficient to demonstrate a relationship between individual vocal characteristics and the preferred delay time of reflections. In this study, to investigate the singer's preferred acoustics with a change in singing style, subjects were asked to perform non-plosive, non-fricative text for the lyrics, using exclusively “la” syllables ( melisma singing). A resulting shift in preferred time delay was observed. The extent of the shift in preferred reflection time delays is shown to be directly related to the minima of the effective duration of the running autocorrelation function calculated from each singer's voice. Singers were also subjected to training, to assist in identifying sound fields. After training, the average preferred delay time of the reflection did not change, but the statistical variability of the singer's subjective rating of the sound fields was strongly reduced.

A STUDY OF THE BLENDING OF VOCAL MUSIC WITH THE SOUND FIELD BY DIFFERENT SINGING STYLES

The initial time delay Δt 1 between the direct sound and the first reflection and the subsequent reverberation time T sub are usually fixed within a given space. Thus, concert halls do not have ideal conditions for all forms of music. It has been shown that the most preferred conditions for both listeners and performers are determined by the minimum value of the effective duration of the running autocorrelation function (ACF) of sound signals, ( τ e ) min . To determine the suitability of vocal music for a given sound field, ( τ e ) min of vocal music was analyzed, after recording five solo singers (tenor) in an anechoic room. The results showed that ( τ e ) min of the ACF of a voice source, which is closely related to the two temporal factors of the sound field, varies with singing style. A significant finding is that the values of ( τ e ) min of sound signals for falsetto and medium falsetto are significantly longer than that for operatic singing.

On the subjective responses based on the auditory-brain model in relation to the factors extracted from the interaural cross-correlation mechanism and the auto-correlation mechanism of sound fields

Based on the auditory-brain model for subjective responses, the author investigated the relationship between several important subjective attributes of the sound field and the orthogonal factors that are extracted from both of the interaural cross-correlation function and the autocorrelation function. There are three significant findings. (1) The apparent source width (ASW) is described by two orthogonal spatial factors: the IACC and the width of the interaural cross-correlation function (WIACC). The scale value of ASW is formulated by superpositioning two terms: the 3/2 power of the IACC and the 1/2 power of the WIACC. (2) The most preferred delay time of a single reflection for a cello soloist can be calculated by the amplitude of the reflection and the minimum value of the effective duration of the running autocorrelation function of the music motifs played by the cellist. (3) From subjective preference tests by paired-comparison method, for different source locations on the stage in an actual concert hall, the subjective preference can be calculated using four orthogonal factors and the interaural time delay of the IACC (IACC). When the IACC is not obtained at = 0IACC, the preference scores decrease rapidly due to an image shift or an unbalanced sound field. Results agree well with the subjective preference obtained by use of simulated sound fields when = 0IACC. Thesis advisor: Yoichi Ando Copies of this thesis written in English can be obtained from Shin-ichi Sato, Graduate School of Science and Technology, Kobe University, Rokkodai, Nada, Kobe 657-8501 Japan. E-mail address: satos@kobe-u.ac.jp.

DIFFERENTIAL EFFECTS OF NOISE AND MUSIC SIGNALS ON THE BEHAVIOR OF CHILDREN

A theory based on the model of how the auditory–brain system perceive primary sensations is used to explain the differential effects of noise and music signals on the sleep of babies and on the performance of mental tasks by children. In a previous study by Ando and Hattori, [1], it was found that sleeping babies (2–4 months old) whose mothers had begun living in a noisy area before conception or during the first five months of pregnancy did not react to daily aircraft noise but did react to music. In another previous study by Ando et al.[2], the percentage of the pupils in “V-type relaxation” state during an adding task in a quiet living area was much greater when pupils heard music than when they heard noise. These phenomena are explained here by the difference between the temporal factors extracted from the running autocorrelation function of the noise and music signals.

SUBJECTIVE PREFERENCE OF CELLISTS FOR THE DELAY TIME OF A SINGLE REFLECTION IN A PERFORMANCE

To provide knowledge useful in designing the stage enclosure in a concert hall, Nakayama reported the subjective preference of alto-recorder players for sound fields with a single reflection [Acustica54, 217–221 (1984)]. The present study evaluates the subjective preferences, with regard to ease of performance, of five cello soloists for the delay time of a single reflection. The scale values of preference for the delay time of a single reflection were obtained using a paired comparison method, and the results were compared with those for the alto-recorder players and listeners. The scale values of preference for both individuals and for global cellists with regard to the delay time of reflection can be expressed by a single approximate formula, normalizing the delay time by the most-preferred delay time observed for different music motifs. A notable finding is that the most-preferred delay time of a single reflection for each cellist can be calculated from the amplitude of the reflection and the minimum value of the effective duration (τe)minof the running autocorrelation function of the music motifs played by each cellist.

RELATIONSHIP BETWEEN SUBJECTIVE PREFERENCE AND THE ALPHA-BRAIN WAVE IN RELATION TO THE INITIAL TIME DELAY GAP WITH VOCAL MUSIC

Previously, it was reported that the most preferred initial time delay gap [Δt1]pand subsequent reverberation time are described by the minimum value of the effective duration (τe)minof the running autocorrelation function (ACF) of music signals (2 T=2·0 s) (Y. ANDO et al. 1989 Journal of Acoustical Society of America86, 644–649). This paper shows whether this result is supported or not by use of the electro-physiological method. Experiments were performed for sound fields changing the initial time delay gapΔt1 of a single reflection with vocal music as a source signal, which has large changes in runningτe . The results at the time interval when (τe)minof the music is observed reveal that the scale value of subjective preference is closely related to the value of τeof the alpha wave obtained from the left heimsphere.

On the relationship between the subjective preference of sound fields and the α-brain wave in both hemispheres

For the purpose of finding possible brain activity corresponding to the subjective preference of sound fields, an attempt was made to analyze the effective duration (τe) of the running autocorrelation function (ACF) in the α-wave range (8–13 Hz) in both hemispheres. As the sound source, a music motif (Arnold: Sinfornietta of Opus 48, a 5 s piece of the 3rd movement) was selected. When the temporal factor of the sound fields (the delay time of reflection Δt1) is changed, the values of τe in the α wave in the left hemisphere correspond well with the subjective preference, but not in the right hemisphere. A long value of τe in the α wave indicates the preferred condition. When the reverberation time Tsub is changed, the values of τe in the α wave were much longer at the preferred reverberation time [Tsub]p=1.2 s than those at Tsub=0.2 or 6.4 s in the left hemisphere. On the other hand, when the spatial factor of the sound fields (IACC) is changed, the right-hemisphere dominance is achieved by the analyses of the value of τe in the α wave (p<0.001).

The running autocorrelation function of different music signals relating to preferred temporal parameters of sound fields

This paper defines the minimum value of the effective duration of the running autocorrelation function of a music signal used in subjective preference testing. A short duration signal, which may relate to ‘‘psychological present,’’ is more suitable for calculating the most preferred time delay of the single reflection than is the long-time autocorrelation function. This fact is found particularly when passages of Japanese music are being played, which include a wide-range-noise component at the very beginning of the signal and the so-called ‘‘MA.’’ MA is a stationary sound passage without any rapid movement. This paper also shows that, for subsequent reverberation time, a flat frequency response may be in the range of preferred conditions. Such preferred conditions may be calculated by the minimum effective duration of the running autocorrelation function of source signals after passing through the A-weighting network.

Further investigation on concert hall acoustics

Since the book [Y. Ando, Concert Hall Acoustics (Springer, New York, 1985)] was published, further work has been carried out. The results demonstrate that the running autocorrelation function of source signals may better describe preferred conditions of sound fields than does the long‐time autocorrelation function. The minimum value of the effective duration of the running autocorrelation function of music signals for the interval used for the subjective preference judgment corresponds well to the preferred initial time delay gap between the direct sound and the first reflection when the amplitude A1 = 0 dB. This was discovered when Japanese music containing a so‐called “MA,” a stationary sound passage without any rapid movement, was used as a source signal. Also, it was found that a flat frequency characteristic of the subsequent reverberation time is the optimal condition, and that it may be calculated by the minimum effective duration of the running autocorrelation function of the source signal. Several other results will be presented.

On the Concept of an Instantaneous Power Spectrum, and Its Relationship to the Autocorrelation Function

A definition sufficient to define a function of frequency ρ(t,f) which could be regarded as the instantaneous power spectrum was given by Page. This function is not unique, since it may have added to it a complementary function of frequency ρc(t,f) satisfying ∫ −∞∞ρc(T,f)df=0without changing the original signal in any way. It is shown that when a signal is observed by different observers, starting their observations at different times, they will not derive the same ``instantaneous power spectral function,'' and the differences between their results will be complementary functions as defined above. It is well known that the Fourier transform of the autocorrelation function is equal to the square of the magnitude of the spectral function. Analogously, a ``running autocorrelation function'' is defined and it is shown that the Fourier transform of its partial differential coefficient with respect to time is an instantaneous power spectral function.