This paper presents a method of evaluating the average decay rate of total room modes inthe observation frequency range. Attending to the decaying sound process of total roomenergy, {<S²(t)>} No, in a reverberation room occupied with absorption specimen, thetheoretical formula L_N(t) was developed which turned out to be bent considerably onthe logarithmic scale due to the lack of diffusion, keeping however its initial decay rate -4.34 αconstant. Actually measured bent decay {<S²(t)>} No is then, through the approximationby L_N(t), characterized simply by two parameters; T_NO (Initial Decay Time) and Q₃₀ (Curvature Index). The process is termed “PLD (Power-law Decay) approach, ” whichrealizes absorption measurements under the observation of, but without being affected by, the diffusion condition of the room. Thus, mounting the specimen so as to cover the entireroom floor eliminating the edge effect, absorption coefficient for infinite specimen, α_∞, isobtained which eventually corresponds to the random incident absorption coefficient sincethey coincide between rooms with greatly different volumes, i.e., different diffusion conditions.Thus the proposed method PLD somehow differs from the conventional method, especially because it validates direct evaluation of average decay rate independent of the curvature on the decay records.

View full abstract

Errors and scatters in usual absorption measurement in a reverberation room are mainlycaused by non-diffusion condition and edge effect, which unfortunately differ from room toroom. This paper presents a method termed “PLD/Deep-well” which eliminates theseerror sources and finally yields random incident absorption coefficient, α_∞, by two procedure:(1) mounting “Deep-well, ” reflective panel taller than the thickness of the specimen, closely around its perimeter to eliminate the edge effect, and (2) applying “PLD (Power-law decay) approach” which compensates the lack of diffusion including what may be aggravated by procedure (1) through numerical evaluation of the curvature on the space-ensemble average decay process {<S² (t)>} _No, i.e., through direct prediction of total room absorptionfor perfect diffusion. Several α_∞ measurements were made to examine the validity of theproposed method, as well as computer simulation to confirm it from the qualitative viewpoint.The results imply the method gives α_∞ with high precision and reproducibility, independent of the area, shape and location of the specimen, diffusion condition, and theheight of the well H only if H≥H₀+t, where t is the thickness of the specimen and H₀ theminimum required height of the well which turned out about 0.8m for usual reverberation room.

View full abstract

This paper presents a between-word distance (BWD) calculation in a symbolic domain anddiscusses its typical application in estimating the degree of the speech recognition difficultyfor given word sets. The first part of the paper describes a method for the distance calculationwhich employs matching by DP (dynamic programming) on subphonemic segmentsequences to take phonemic-context-dependent characteristics into account. In order to testthe usefulness of the method, two types of word sets are composed using a distance-basedclustering technique. Vocabularies of one type have dense sample distributions while theothers have sparse sample distributions in a BWD sense. Speaker-independent word recognitionis examined for these word sets using a common phone-HMM-based speech recognitiontechnique. We compare the recognition results and the statistical characteristics ofindividual word sets, and present criteria for relative order of the recognition difficulty ofgiven word sets. One criterion using between-word distance distributions of n-nearest neighbor words provides a reasonable index for the recognition difficulty.

View full abstract

The estimation of transfer function and impulse response of a system by the cross spectraltechnique is very popular and most of the FFT analyzer is equipped with that function. Inthe cross spectral technique, the impulse response changes its form by the use of timewindow. But the deformation of the impulse response in the process of estimation is not yetwell known. This paper first theoretically describes why and how the impulse responsechanges its form. According to the theoretical derivation, the deformation of the impulseresponse is illustrated for several commonly used time windows and the results are verified by the computer simulation.

View full abstract

In recent years, traffic noise gives rise to public discussion with the development of transportfacilities. An exhaust noise of engine which is one of the traffic noise is classified intoperiodic one. The most power of a periodic noise to be controlled is concentrated oncertain harmonic components, so that it is effective to control only these harmonic components.When active noise control (ANC) systems are actually implemented, they need amodel of characteristics of secondary path. The difference between the actual characteristicsof secondary path and its model is called modeling error. It is known that a degree ofthe modeling error affects the performance of ANC system directly. In this paper, a newadaptive algorithm to control periodic noise is proposed. The proposed algorithm is robustagainst an observation noise and its computational load is smaller than that of conventionaladaptive algorithm. Also its extended algorithm which has on-line estimate function of adelay component of secondary path is proposed. The performance of these algorithms isevaluated by computer simulations, model experiment using a duct and an experiment using exhaust noise of diesel engine.

View full abstract