diff --git a/tests/README.md b/tests/README.md
index bedb35a7..0b725dba 100644
--- a/tests/README.md
+++ b/tests/README.md
@@ -92,3 +92,26 @@ Sound System 11         | IAMF                     | L3, R3, C, LFE, Ltf3, Rtf3,
 Sound System 12         | IAMF                     | C
 Sound System 13         | IAMF                     | FL, FR, FC, LFE, BL, BR, FLc, FRc, SiL, SiR, TpFL, TpFR, TpBL, TpBR, TpSiL, TpSiR
 Binaural Layout         | IAMF                     | L2, R2
+
+# Verification
+
+For test cases using Opus or AAC codecs, the average PSNR value must exceed 30, and for the other codecs, an average PSNR value exceeding 80 is considered PASS.
+You can use `psnr_calc.py` file to calculate PSNR between reference and generated output.
+
+- How to use `psnr_calc.py` script:
+    ```
+    python psnr_calc.py
+      --dir <directory path containing the target file and reference file>
+      --target <target wav file name>
+      --ref <reference wav file name>
+    ```
+
+- Calculate PSNR values of multiple wav files
+
+    Multiple files can be entered as `::`
+
+    ```
+    Example:
+
+    python psnr_calc.py --dir . --target target1.wav::target2.wav --ref ref1.wav::ref2.wav
+    ```
\ No newline at end of file
diff --git a/tests/psnr_calc.py b/tests/psnr_calc.py
new file mode 100644
index 00000000..06f737fe
--- /dev/null
+++ b/tests/psnr_calc.py
@@ -0,0 +1,143 @@
+import argparse
+import wave
+import os
+import scipy.io.wavfile as wavfile
+import numpy as np
+import math
+
+parser = argparse.ArgumentParser(description="PSNR verification script")
+parser.add_argument(
+    "--dir",
+    type=str,
+    required=True,
+    help="decoder verification wav output directory",
+)
+parser.add_argument(
+    "--target",
+    type=str,
+    required=True,
+    help="decoder verification wav output file. Multiple files can be entered as ::. (ex - test1.wav::test2.wav)",
+)
+parser.add_argument(
+    "--ref",
+    type=str,
+    required=True,
+    help="decoder verification PSNR evaluation reference file. Multiple files can be entered as ::. (ex - test1.wav::test2.wav)",
+)
+args = parser.parse_args()
+
+
+def get_sampwdith(path):
+    with wave.open(path, "rb") as wf:
+        sampwidth_bytes = wf.getsampwidth()
+    return sampwidth_bytes
+
+
+def calc_psnr(ref_signal, signal, sampwidth_bytes):
+    assert (
+        sampwidth_bytes > 1
+    ), "Supports sample format: [pcm_s16le, pcm_s24le, pcm_s32le]"
+    max_value = pow(2, sampwidth_bytes * 8) - 1
+
+    # To prevent overflow
+    ref_signal = ref_signal.astype("int64")
+    signal = signal.astype("int64")
+
+    mse = np.mean((ref_signal - signal) ** 2, axis=0, dtype="float64")
+
+    psnr_list = list()
+
+    # To support mono signal
+    num_channels = 1 if ref_signal.shape[1:] == () else ref_signal.shape[1]
+    for i in range(num_channels):
+        mse_value = mse[i] if num_channels > 1 else mse
+        if mse_value == 0:
+            print(f"ch#{i} PSNR: inf")
+        else:
+            psnr_value = 10 * math.log10(max_value**2 / mse_value)
+            psnr_list.append(psnr_value)
+            print(f"ch#{i} PSNR: {psnr_value} dB")
+
+    return -1 if len(psnr_list) == 0 else sum(psnr_list) / len(psnr_list)
+
+
+target_file_list = args.target.split("::")
+ref_file_list = args.ref.split("::")
+
+tc_number_list = []
+psnr_list = []
+for file_idx in range(len(target_file_list)):
+    target_file = target_file_list[file_idx]
+    ref_file = ref_file_list[file_idx]
+    print(
+        "[%d] PSNR evaluation: compare %s with %s"
+        % (file_idx, target_file, ref_file)
+    )
+    tc_number_list.append(file_idx)
+    try:
+        ref_filepath = os.path.join(
+            os.path.dirname(os.path.abspath(__file__)), args.dir, ref_file
+        )
+        target_filepath = os.path.join(
+            os.path.dirname(os.path.abspath(__file__)), args.dir, target_file
+        )
+
+        ref_samplerate, ref_data = wavfile.read(ref_filepath)
+        target_samplerate, target_data = wavfile.read(target_filepath)
+
+        ref_sampwdith_bytes = get_sampwdith(ref_filepath)
+        target_sampwidth_bytes = get_sampwdith(target_filepath)
+
+        # Check sampling rate
+        if not (ref_samplerate == target_samplerate):
+            print(ref_file, " / ", target_file)
+            raise Exception(
+                "Sampling rate of reference file and comparison file are different."
+            )
+
+        # Check number of channels
+        if not (ref_data.shape[1:] == target_data.shape[1:]):
+            raise Exception(
+                "Number of channels of reference file and comparison file are different."
+            )
+
+        # Check number of samples
+        if not (ref_data.shape[0] == target_data.shape[0]):
+            print(ref_file, " / ", target_file)
+            raise Exception(
+                "Number of samples of reference file and comparison file are different."
+            )
+
+        # Check bit depth
+        if not (ref_sampwdith_bytes == target_sampwidth_bytes):
+            print(ref_file, " / ", target_file)
+            raise Exception(
+                "Bit depth of reference file and comparison file are different."
+            )
+
+        average_psnr = calc_psnr(ref_data, target_data, ref_sampwdith_bytes)
+        if average_psnr != -1:
+            print("average PSNR: %.15f" % (average_psnr))
+            psnr_list.append(average_psnr)
+        else:
+            print("average PSNR: %.15f" % (100))
+            psnr_list.append(100)
+    except Exception as err:
+        print(str(err))
+        psnr_list.append(0)
+    print("")
+
+# print result
+print(
+    "\n\n\n[Result] - (If the OPUS or AAC codec has a over avgPSNR 30, it is considered PASS. Other codecs must be over avgPSNR 80.)"
+)
+for i in range(len(tc_number_list)):
+    print(
+        "TC#%d : %.3f (compare %s with %s)"
+        % (
+            tc_number_list[i],
+            round(psnr_list[i], 3),
+            target_file_list[i],
+            ref_file_list[i],
+        )
+    )