Step 1: MTF to Apparent SNR

Each MTF value m(Fm,k) is converted to an apparent signal-to-noise ratio:

SNRapp(Fm,k) = 10 · log₁₀ [ m / (1 - m) ] (dB)

This logarithmic conversion maps MTF 0.5 → 0 dB, MTF 0.99 → +20 dB, MTF 0.01 → -20 dB. The apparent SNR is then truncated to the range -15 dB to +15 dB (Clause 5.3). Values outside this range contribute no additional information about intelligibility.

Step 2: Modulation Transmission Index per Band

The Transmission Index for each octave band k is the average of the truncated apparent SNRs across all modulation frequencies, normalized to a 0–1 scale:

TI(k) = [ (1/N) · Σ SNRapp(Fm,k) + 15 ] / 30

Where N is 14 for full STI or 2 for STIPA. TI(k) = 0 corresponds to apparent SNR of -15 dB in all modulation frequencies; TI(k) = 1 corresponds to +15 dB.

Step 3: Band Weighting (Table 4)

The 7 octave-band TI values are combined using male-speech weighting factors from Table 4:

Band (Hz)	125	250	500	1k	2k	4k	8k
Weight α	0.085	0.127	0.230	0.233	0.309	0.224	0.173

Note the weights sum to more than 1.0 — this is intentional, as the redundancy correction in Step 4 subtracts the inter-band correlation contribution.

Step 4: Redundancy Correction (Table 5)

Adjacent octave bands share correlated speech information. The redundancy correction subtracts this overlap:

STI = Σ α(k)·TI(k) - Σ β(k,k+1)·√[TI(k)·TI(k+1)]

The β values (Table 5) account for inter-band correlation: 0.085, 0.078, 0.065, 0.011, 0.047, 0.095 for adjacent band pairs. This correction typically reduces the raw weighted sum by 0.02–0.05.

Step 5: Final STI

The result is a single number from 0 to 1, mapped to the qualitative scale: Bad (<0.30), Poor (0.30-0.45), Fair (0.45-0.60), Good (0.60-0.75), Excellent (>0.75). The CIS (Common Intelligibility Scale) provides a percentage-like alternative: CIS = 1 + log₁₀(STI).

Simplified STI from RT60 + SNR

When only RT60 and SNR are known (no direct measurement), STI can be estimated using the theoretical MTF formula for exponential decay plus noise. This is less accurate but useful for design predictions.

SonaVyx Calculation

The STI tool performs the complete calculation pipeline. Pro mode shows intermediate values at each step — per-band apparent SNR, TI values, weighted contribution, redundancy correction — for full transparency. See masking corrections for the auditory masking model applied before Step 1.