diff --git a/src/util.jl b/src/util.jl
index 00fedb8b0..90b781e34 100644
--- a/src/util.jl
+++ b/src/util.jl
@@ -4,6 +4,7 @@ import Base: *
 import LinearAlgebra.BLAS
 using LinearAlgebra: mul!
 using FFTW
+using Statistics: mean
 
 export  hilbert,
         fftintype,
@@ -162,11 +163,18 @@ Convert an amplitude ratio to dB (decibel), or ``20
 amp2db(a::Real) = 20*log10(a)
 
 """
-    rms(s)
+    rms(s; dims)
 
-Return the root mean square of signal `s`.
+Return the root mean square (rms) of signal `s`. Optional keyword parameter
+`dims` can be used to specify the dimensions along which to compue the rms.
 """
-rms(s::AbstractArray{T}) where {T<:Number} = sqrt(sum(abs2, s)/length(s))
+function rms(s::AbstractArray{T}; dims=:) where {T<:Number}
+    if dims === (:)
+        return sqrt(mean(abs2, s))
+    else
+        return sqrt.(mean(abs2, s; dims=dims))
+    end
+end
 
 """
     rmsfft(f)
diff --git a/test/util.jl b/test/util.jl
index b302a2744..ed6f194e7 100644
--- a/test/util.jl
+++ b/test/util.jl
@@ -83,6 +83,9 @@ end
     n = (5,6)
     for x in ( randn(n), randn(n)+randn(n)im )
         @test rms(x) ≈ sqrt(mean(abs.(x).^2))
+        @test rms(x; dims=1) ≈ sqrt.(mean(abs.(x).^2; dims=1))
+        @test rms(x; dims=2) ≈ sqrt.(mean(abs.(x).^2; dims=2))
+        @test rms(x; dims=1:2) ≈ sqrt.(mean(abs.(x).^2; dims=1:2))
         @test rmsfft(fft(x)) ≈ rms(x)
     end # for
 end