Speech interference level (SIL)

The speech interference level was developed as a simple derivative of the articulation index.1,16 It uses octave rather than ⅓ octave analysis of background noise.

The SIL changed with the application of the internationally adopted octave bands and became known as the preferred speech interference level (PSIL).16

The PSIL was the arithmetic mean of the sound pressure level in the 3 octave bands centred on 500, 1000, 2000 Hz. However the PSIL was never standardized and in 1977, the American National Standards Institute standardized the speech interference level.19 This was the same as the PSIL with the addition of the frequencies centred on 4000 Hz. At present, the American National Standards Institute has recommended a speech interference level centred on frequencies 500, 1000, 2,000, 4000 Hz as providing the best estimate of masking noise. This is often abbreviated to SIL (0.5, 1.0, 2.0, 4.0).1

Lazarus7 described the simplified prerequisites for the SIL method as:

• The speech level at 1 metre distances (LSA,1m)is determined by the vocal effect. The speech level can be determined by the sound pressure level at the position of the speaker.

• Sound propagation corresponds to that of the free field.

• The sound level determines the speech interference level or the approximate A- weighted sound level.

• The relationship between the speech interference level and distance between speaker and recipient is defined by a given speech intelligibility or signal-to-noise ratio.

• A higher signal to noise ratio must be used for higher vocal effect because speech intelligibility decreases at higher speech levels due to distortions and vocal strain. The necessary signal-to-noise ratios for the higher vocal effects are given in Table 4-3 (below).

Table 4-3 Necessary signal-to-noise ratios for different vocal effects

Vocal effect Signal-to-noise ratio (SNR) (dB)

Very loud 3

Shout 5

Maximum 8

Source: Lazarus7

The time-average sound level is used for fluctuating sound. The A-weighted sound level is used as an estimation value for the speech interference level (LSIL). The A- weighted sound level approximates the speech-interference level LSIL+ 8 dB for most sounds as shown in Figure 4-5 below:

Figure 4-5 Relationship of speech interference level to distance and vocal effect in the free field.

Source: Lazarus7, Berglund and Lindvall1

Figure 4-5 (above) shows the relationship between the speech interference level (LSIL) /A-weighted noise level (LNA) and the speaker-listener distance (r) for satisfactory speech intelligibility for a vocal effect: relaxed to maximum shouting. The normal vocal effect is a speech level at 1 M (LSA,1M) = 60 dB.

Satisfactory speech intelligibility means that for the vocal effect relaxed to loud, the articulation index (AI) = 0.45 or the SNR = 3 dB. For the higher vocal effects to achieve satisfactory speech intelligibility, the signal-to-noise ratios are given in Table 4- 3 (above).

There have been some changes made to the speech interference level with some minor adjustments. Kryter17 gave the speech interference levels of steady continuous noise at

which reliable communication can just be perceived (face-to-face). These were based on the original pre-standardized SILs. Studies have found that the standardized SILs are approximately 1 dB higher than the original SIL and 2 dB lower than the PSILs, which is well within the margin of error of sound level measurements. However it is reasonable to correct the original SILs as given in Kryter to standardized SILs which are given in Table 4-4 based on the following criteria:19

• Original SIL + 1 dB = Standardized SIL

• The standardized SILs are approximately 2dB lower than the PSILs

Table 4-4 Standardized speech interference levels of steady continuous noise at which just reliable communication can occur.

Speech interference level (standardized) Separation

distance

Normal effort Raised Very loud Shouting

0.15m *72 78 84 90 0.3m 66 72 78 84 0.6m 60 66 72 78 1.2m 54 60 66 72 1.8m 50 56 62 68 3.7m 44 50 56 62

* The PSIL = 74 i.e. 2 dB higher than the standardized SIL as explained above.

Adapted from Kryter17 and May16

May16 has suggested a correction for adult females where 5 dB is deduced for females (and possibly children). As the preferred speech interference levels are approximately 2 dB higher than the standardised figures, it is reasonable to assume the same deduction of 5 dB for the standardized speech interference levels (Table 4-4), due to the obvious level of uncertainty in such perceived measurements.

May16 also indicates an approximate relationship between the preferred speech interference level (PSIL) and most other sounds other than speech. The background noise is approximately equal to the PSIL + 7dB. This varies slightly for what Lazarus7

with a SIL of 70 dB, the level of other competing noise is likely to be in the region of 78 dB.

May16 gave an example of this usage but used what is now described as the non- standardized 3-band procedure5 based on the PSIL which excluded the 4000 Hz frequency band. If this method is to be used, Levitt and Webster have instructed that the frequencies must be included as shown to avoid confusion. SIL (0.5, 1, 2). The standardized 4-band method includes the 4000 Hz band). Both methods are demonstrated in the worked example (adapted from May) below:

Determine if a noise is acceptable for speech on a bus if the octave band analysis of noise received at a passenger’s ear on public transport is as follows:

Octave-band centre frequencies (Hz) 125 250 500 1000 2000 4000 Sound pressure level (dB) 90 87 80*+ 75*+ 70*+ 67*

Determine if this spectrum is acceptable for speech?

Answer using the Standardized SIL (4-band) method:

• Ignore frequencies outside 500 – 4000 Hz target range.

• Average the sound pressure levels* in the target range.

i.e. SIL = ) 73dB 4 67 70 75 80 ( + + + =

Answer using the SIL (0.5,1,2) (3 band) method

• Ignore frequencies outside 500 - 2000Hz target range

• Average the sound pressure levels+ in the target range. ie SIL (0.5,1,2) = ) 75dB 3 70 75 80 ( + + =

If on a bus or train, we would expect to be able to talk in a normal voice to people one seat away (< 1metre). At 0.6 metres, the SIL = 60 dB for a man and (60-5) = 55 dB for a woman/child (assuming the same correction for SIL as for PSIL). Since the calculated SIL of 73 dB [or SIL (0.5,1,2) of 75 dB], both exceed the tabulated values of 60 dB in Table 4-4 (and 55 dB when corrected for females), this sound is unacceptably loud for close conversation.16

Lazarus7 described the different ways the response function of the room can influence the speech sound, the equivalent noise ratio and therefore the speech intelligibility. These are:

• The influence of reverberation on speech and therefore intelligibility increases with the distance between speaker and listener.

• A reverberation time of no more than 0.5 seconds can slightly improve speech intelligibly for those with normal hearing but greater reverberation times decrease intelligibility.

Work has been done to compare speech intelligibly measurements in rooms using the three procedures, articulation index, speech transmission index and A-weighted signal- to-noise ratio. Rooms including classrooms with different reverberation times and signal receiver distances were used. Statistical analysis of all three procedures showed them to be relatively close in prediction of speech intelligibility.20, 21

There is no simple formula for predicting the relationships for communication indoors. Figure 4-5 (above) represents relationships in the free field (outdoors) with the application of the inverse square law. Berglund and Lindvall1 indicate that Figure 4-5 can be used to determine permissible noise levels for a specific distance up to 2 metres and as an estimate up to 8 metres if the reverberation time is lower than 2 seconds. Standards can also be set on the basis of average sound pressure levels found to be acceptable in similar settings from past experience.1